
#pip install scikit-plot


#pip install shap


import math
import shap
import numpy as np
import pandas as pd
from scipy import stats
import matplotlib.pyplot as plt
import seaborn as sns
import scikitplot as skplt
import plotly.subplots as sp
import plotly.graph_objects as go
import plotly.express as px
import tensorflow as tf

from matplotlib import cm

from sklearn.preprocessing import LabelEncoder, StandardScaler
from sklearn.linear_model import LinearRegression, Lasso
from sklearn.pipeline import Pipeline
from sklearn.ensemble import RandomForestRegressor, RandomForestClassifier
from sklearn.tree import DecisionTreeRegressor, DecisionTreeClassifier
from sklearn.svm import SVR
from sklearn.naive_bayes import GaussianNB
from sklearn.neighbors import KNeighborsClassifier
from sklearn.model_selection import train_test_split, cross_val_score, GridSearchCV
from sklearn.metrics import roc_curve, auc, r2_score, mean_squared_error, accuracy_score, f1_score

from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import Dense, Dropout, BatchNormalization
from tensorflow.keras.utils import to_categorical
from tensorflow.keras.callbacks import EarlyStopping

from bs4 import BeautifulSoup
import requests

WARNING:tensorflow:From C:\Users\sebmo\anaconda3\lib\site-packages\keras\src\losses.py:2976: The name tf.losses.sparse_softmax_cross_entropy is deprecated. Please use tf.compat.v1.losses.sparse_softmax_cross_entropy instead.


data = pd.read_csv("OnlineNewsPopularity.csv")
data.head()


data.shape

(39644, 61)


data.isna().sum()

url                              0
 timedelta                       0
 n_tokens_title                  0
 n_tokens_content                0
 n_unique_tokens                 0
                                ..
 title_subjectivity              0
 title_sentiment_polarity        0
 abs_title_subjectivity          0
 abs_title_sentiment_polarity    0
 shares                          0
Length: 61, dtype: int64


data.columns

Index(['url', ' timedelta', ' n_tokens_title', ' n_tokens_content',
       ' n_unique_tokens', ' n_non_stop_words', ' n_non_stop_unique_tokens',
       ' num_hrefs', ' num_self_hrefs', ' num_imgs', ' num_videos',
       ' average_token_length', ' num_keywords', ' data_channel_is_lifestyle',
       ' data_channel_is_entertainment', ' data_channel_is_bus',
       ' data_channel_is_socmed', ' data_channel_is_tech',
       ' data_channel_is_world', ' kw_min_min', ' kw_max_min', ' kw_avg_min',
       ' kw_min_max', ' kw_max_max', ' kw_avg_max', ' kw_min_avg',
       ' kw_max_avg', ' kw_avg_avg', ' self_reference_min_shares',
       ' self_reference_max_shares', ' self_reference_avg_sharess',
       ' weekday_is_monday', ' weekday_is_tuesday', ' weekday_is_wednesday',
       ' weekday_is_thursday', ' weekday_is_friday', ' weekday_is_saturday',
       ' weekday_is_sunday', ' is_weekend', ' LDA_00', ' LDA_01', ' LDA_02',
       ' LDA_03', ' LDA_04', ' global_subjectivity',
       ' global_sentiment_polarity', ' global_rate_positive_words',
       ' global_rate_negative_words', ' rate_positive_words',
       ' rate_negative_words', ' avg_positive_polarity',
       ' min_positive_polarity', ' max_positive_polarity',
       ' avg_negative_polarity', ' min_negative_polarity',
       ' max_negative_polarity', ' title_subjectivity',
       ' title_sentiment_polarity', ' abs_title_subjectivity',
       ' abs_title_sentiment_polarity', ' shares'],
      dtype='object')


column_names = []
for column in data.columns:
  column_names.append(column.lstrip())
data.columns = column_names


data.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 39644 entries, 0 to 39643
Data columns (total 61 columns):
 #   Column                         Non-Null Count  Dtype  
---  ------                         --------------  -----  
 0   url                            39644 non-null  object 
 1   timedelta                      39644 non-null  float64
 2   n_tokens_title                 39644 non-null  float64
 3   n_tokens_content               39644 non-null  float64
 4   n_unique_tokens                39644 non-null  float64
 5   n_non_stop_words               39644 non-null  float64
 6   n_non_stop_unique_tokens       39644 non-null  float64
 7   num_hrefs                      39644 non-null  float64
 8   num_self_hrefs                 39644 non-null  float64
 9   num_imgs                       39644 non-null  float64
 10  num_videos                     39644 non-null  float64
 11  average_token_length           39644 non-null  float64
 12  num_keywords                   39644 non-null  float64
 13  data_channel_is_lifestyle      39644 non-null  float64
 14  data_channel_is_entertainment  39644 non-null  float64
 15  data_channel_is_bus            39644 non-null  float64
 16  data_channel_is_socmed         39644 non-null  float64
 17  data_channel_is_tech           39644 non-null  float64
 18  data_channel_is_world          39644 non-null  float64
 19  kw_min_min                     39644 non-null  float64
 20  kw_max_min                     39644 non-null  float64
 21  kw_avg_min                     39644 non-null  float64
 22  kw_min_max                     39644 non-null  float64
 23  kw_max_max                     39644 non-null  float64
 24  kw_avg_max                     39644 non-null  float64
 25  kw_min_avg                     39644 non-null  float64
 26  kw_max_avg                     39644 non-null  float64
 27  kw_avg_avg                     39644 non-null  float64
 28  self_reference_min_shares      39644 non-null  float64
 29  self_reference_max_shares      39644 non-null  float64
 30  self_reference_avg_sharess     39644 non-null  float64
 31  weekday_is_monday              39644 non-null  float64
 32  weekday_is_tuesday             39644 non-null  float64
 33  weekday_is_wednesday           39644 non-null  float64
 34  weekday_is_thursday            39644 non-null  float64
 35  weekday_is_friday              39644 non-null  float64
 36  weekday_is_saturday            39644 non-null  float64
 37  weekday_is_sunday              39644 non-null  float64
 38  is_weekend                     39644 non-null  float64
 39  LDA_00                         39644 non-null  float64
 40  LDA_01                         39644 non-null  float64
 41  LDA_02                         39644 non-null  float64
 42  LDA_03                         39644 non-null  float64
 43  LDA_04                         39644 non-null  float64
 44  global_subjectivity            39644 non-null  float64
 45  global_sentiment_polarity      39644 non-null  float64
 46  global_rate_positive_words     39644 non-null  float64
 47  global_rate_negative_words     39644 non-null  float64
 48  rate_positive_words            39644 non-null  float64
 49  rate_negative_words            39644 non-null  float64
 50  avg_positive_polarity          39644 non-null  float64
 51  min_positive_polarity          39644 non-null  float64
 52  max_positive_polarity          39644 non-null  float64
 53  avg_negative_polarity          39644 non-null  float64
 54  min_negative_polarity          39644 non-null  float64
 55  max_negative_polarity          39644 non-null  float64
 56  title_subjectivity             39644 non-null  float64
 57  title_sentiment_polarity       39644 non-null  float64
 58  abs_title_subjectivity         39644 non-null  float64
 59  abs_title_sentiment_polarity   39644 non-null  float64
 60  shares                         39644 non-null  int64  
dtypes: float64(59), int64(1), object(1)
memory usage: 18.5+ MB


def binary_columns(df):
    cols_binaires = []

    for col in data.columns:
          if len(data[col].value_counts()) == 2:
              cols_binaires.append(col)

    return cols_binaires


colonnes_binaires_df = data.filter(items=binary_columns(data))
colonnes_binaires_df.describe()


data.drop(columns=binary_columns(data), inplace=False).hist(bins=50, figsize= (25,16))
plt.show()


sns.boxplot(x = data['shares'])
plt.title("Boxplot of the Shares' distribution")
#plt.xlim(0,15000)
plt.show()


print("Quantile 90% :",data['shares'].quantile(0.9))
print("Number of rows above 90% :", data[data['shares']>data['shares'].quantile(0.9)].shape[0]," so about ",round(data[data['shares']>data['shares'].quantile(0.9)].shape[0]*100/len(data),2),"% of the dataset.\n")
print("Quantile 75% :",data['shares'].quantile(0.75))
print("Number of rows above 75% :", data[data['shares']>data['shares'].quantile(0.75)].shape[0]," so about ",round(data[data['shares']>data['shares'].quantile(0.75)].shape[0]*100/len(data),2),"% of the dataset.\n")
print("Quantile 1% :",data['shares'].quantile(0.01))
print("Number of rows under 1% :", data[data['shares']<data['shares'].quantile(0.01)].shape[0]," so about ",round(data[data['shares']<data['shares'].quantile(0.01)].shape[0]*100/len(data),2),"% of the dataset.")

Quantile 90% : 6200.0
Number of rows above 90% : 3946  so about  9.95 % of the dataset.

Quantile 75% : 2800.0
Number of rows above 75% : 9630  so about  24.29 % of the dataset.

Quantile 1% : 381.0
Number of rows under 1% : 394  so about  0.99 % of the dataset.


fig, (ax1,ax2) = plt.subplots(1,2,figsize=(12,6))
sns.boxplot(x = data['shares'], ax=ax1)
ax1.set_xlim(0,data['shares'].quantile(0.75))
ax1.set_title("Without outliers > Quantile 75%")
sns.boxplot(x = data['shares'], ax=ax2)
ax2.set_xlim(0,data['shares'].quantile(0.9))
ax2.set_title("Without outliers > Quantile 90%")
plt.show()


Q1 = data['shares'].quantile(0.25)
Q3 = data['shares'].quantile(0.75)
IQR = Q3-Q1
print("Value at the quantile 25%:",Q1)
print("Value at the quantile 75%:",Q3)
print("Value of the difference between these two quantiles:" ,IQR)

Value at the quantile 25%: 946.0
Value at the quantile 75%: 2800.0
Value of the difference between these two quantiles: 1854.0


lower_lim_shares = Q1 - (1.5 * IQR)
upper_lim_shares = Q3 + 1.5 * IQR
print("Lower limit :",lower_lim_shares)
print("Upper limit :",upper_lim_shares)
upper_quantile_shares = data[data['shares']>upper_lim_shares].shape[0]/len(data)
print("\nThe upper quantile for 'upper_lim' :",100-(upper_quantile_shares*100),"%")
print("Number of rows above upper_quantile :", data[data['shares']>data['shares'].quantile(1-upper_quantile_shares)].shape[0])

Lower limit : -1835.0
Upper limit : 5581.0

The upper quantile for 'upper_lim' : 88.54555544344667 %
Number of rows above upper_quantile : 4541


sns.boxplot(x = data['shares'])
plt.xlim(0,data['shares'].quantile(1-upper_quantile_shares))
plt.title("Without outliers > Quantile 88,55%")
plt.show()


z = np.abs(stats.zscore(data['shares']))
threshold = 3 #classic threshold, but even if 1, the outliers are still too present
outliers = data[z > threshold]
without_outliers = data[z <= threshold]
print("Outliers size (number of rows) :",outliers.shape[0], "so about",round(outliers.shape[0]*100/len(data),2),"% of the dataset.")
sns.boxplot(x=without_outliers['shares'])
plt.title("Without outliers using Z-score and threshold/standard deviation = 3")
plt.show()

Outliers size (number of rows) : 308 so about 0.78 % of the dataset.


data = data[data['shares']<=data['shares'].quantile(1-upper_quantile_shares)]
data.shape

(35103, 61)


data.drop(columns=binary_columns(data), inplace=False).hist(bins=50, figsize= (25,16))
plt.show()


def IQRrule(cols):
    parameters={}
    for col in cols :
        Q1=data[col].quantile(0.25)
        Q3=data[col].quantile(0.75)
        IQR=Q3-Q1
        lower_lim = round((Q1 - (1.5 * IQR)),2)
        upper_lim = round((Q3 + (1.5 * IQR)),2)

        decimal = str(upper_lim).split('.')[1][0]
        if int(decimal) >=5:
          upper_lim = math.ceil(upper_lim)
        else:
          upper_lim = math.floor(upper_lim)

        decimal = str(lower_lim).split('.')[1][0]
        if int(decimal) >=5:
          lower_lim = math.floor(lower_lim)
        else :
          lower_lim = math.ceil(lower_lim)
        print(f"For '{col}' :\nLower limit :",lower_lim)
        print("Upper limit :",upper_lim)
        upper_quantile = data[data[col]>upper_lim].shape[0]/len(data)
        lower_quantile = data[data[col]<lower_lim].shape[0]/len(data)
        print("The upper quantile for 'upper_lim' :",100-(upper_quantile*100),"%")
        print("The upper quantile for 'lower_lim' :",(lower_quantile*100),"%\n")

        parameters[col]=[upper_lim,lower_lim,upper_quantile,lower_quantile]
    return parameters


def DropColumn(df, cols):
    for col in cols:
        if col not in df.columns:
            cols.remove(col)
            print(f"{col} wasn't in the DataFrame.")
    if len(cols)==0:
        print('Nothing has been deleted.')
    else:
        df.drop(cols, axis=1, inplace=True)


token_columns = ['n_tokens_title',
                 'n_tokens_content',
                 'n_unique_tokens',
                 'n_non_stop_words',
                 'n_non_stop_unique_tokens',
                 'average_token_length']


data[token_columns].describe()


print(data['n_non_stop_words'].value_counts().head(10))
print(data['n_non_stop_words'].unique())

0.0    970
1.0    122
1.0    121
1.0    120
1.0    117
1.0    114
1.0    113
1.0    113
1.0    113
1.0    112
Name: n_non_stop_words, dtype: int64
[0.99999999 0.99999999 0.99999999 ... 1.         1.         1.        ]


DropColumn(data,['n_non_stop_words'])
token_columns.remove('n_non_stop_words')


for col in token_columns:
    fig, (ax1,ax2) = plt.subplots(1,2,figsize=(20,6))
    ax1.scatter(data[col],data["shares"])
    ax2.hist(data[col],bins='auto')
    ax1.set_xlabel(col)
    ax1.set_ylabel("Shares")
    ax1.set_title(f"Shares compared to {col}")
    ax2.set_xlabel(col)
    ax2.set_ylabel("Frequency")
    ax2.set_title(f"Distribution of {col}")
    plt.show()


plt.figure(figsize=(16,6))
sns.heatmap(data[token_columns].corr(),cmap="coolwarm",annot=True)
plt.show()


DropColumn(data,['n_non_stop_unique_tokens'])
token_columns.remove('n_non_stop_unique_tokens')


token_columns_no_binary_no_normal_distribution = ['n_tokens_title','n_tokens_content','average_token_length']
IQRparameters = IQRrule(token_columns_no_binary_no_normal_distribution)
IQR_n_tokens_title = IQRparameters['n_tokens_title']
IQR_n_tokens_content = IQRparameters['n_tokens_content']
IQR_avg_token_length = IQRparameters['average_token_length']

For 'n_tokens_title' :
Lower limit : 4
Upper limit : 17
The upper quantile for 'upper_lim' : 99.93162977523289 %
The upper quantile for 'lower_lim' : 0.01709255619177848 %

For 'n_tokens_content' :
Lower limit : -448
Upper limit : 1413
The upper quantile for 'upper_lim' : 95.17705039455318 %
The upper quantile for 'lower_lim' : 0.0 %

For 'average_token_length' :
Lower limit : 3
Upper limit : 5
The upper quantile for 'upper_lim' : 87.73039341366835 %
The upper quantile for 'lower_lim' : 2.763296584337521 %


pd.set_option('display.max_rows', None)
print("Videos :\n",data['num_videos'].value_counts(),"\n")
print("Images :\n",data['num_imgs'].value_counts())

Videos :
 0.0     22583
1.0      8163
2.0      1862
3.0       460
11.0      226
4.0       213
10.0      184
21.0      158
5.0       136
26.0      113
6.0        99
9.0        88
7.0        80
16.0       79
12.0       69
8.0        68
25.0       62
13.0       61
15.0       58
14.0       55
20.0       51
17.0       46
27.0       26
18.0       22
22.0       19
31.0       14
19.0       14
33.0       13
23.0       10
28.0        9
24.0        9
50.0        5
73.0        5
32.0        4
34.0        4
74.0        4
51.0        4
35.0        3
30.0        3
29.0        3
36.0        2
48.0        2
58.0        2
59.0        2
75.0        2
42.0        2
38.0        2
65.0        1
66.0        1
53.0        1
46.0        1
Name: num_videos, dtype: int64 

Images :
 1.0      16698
0.0       5942
2.0       2705
3.0       1092
11.0       930
10.0       711
4.0        591
6.0        507
12.0       480
9.0        455
13.0       431
8.0        431
5.0        415
15.0       404
7.0        362
14.0       327
16.0       326
20.0       260
21.0       208
17.0       182
18.0       154
19.0       147
25.0       138
22.0       114
26.0        96
23.0        93
30.0        83
35.0        69
32.0        69
24.0        62
31.0        59
51.0        58
33.0        56
34.0        35
28.0        35
27.0        33
29.0        32
36.0        29
37.0        23
50.0        23
61.0        20
52.0        18
60.0        14
38.0        13
42.0        12
47.0        11
49.0        10
39.0         9
41.0         8
45.0         8
48.0         8
55.0         8
64.0         7
53.0         7
40.0         7
43.0         5
98.0         5
44.0         5
54.0         4
67.0         4
46.0         4
56.0         4
92.0         4
99.0         4
62.0         4
71.0         3
80.0         3
75.0         3
100.0        3
101.0        2
76.0         2
63.0         2
57.0         1
111.0        1
65.0         1
108.0        1
59.0         1
83.0         1
84.0         1
68.0         1
93.0         1
73.0         1
58.0         1
69.0         1
66.0         1
128.0        1
91.0         1
70.0         1
79.0         1
Name: num_imgs, dtype: int64


pd.set_option('display.max_columns', None)
print("For videos :\n",data[data['num_videos']>=10]['url'][:15])
print("\nFor images :\n",data[data['num_imgs']>=10]['url'][:15])

For videos :
 99     http://mashable.com/2013/01/09/blue-umbrella-t...
126         http://mashable.com/2013/01/09/gopro-videos/
181    http://mashable.com/2013/01/10/al-roker-pooped...
190      http://mashable.com/2013/01/10/cats-in-pajamas/
206    http://mashable.com/2013/01/10/ferrell-gosling...
232        http://mashable.com/2013/01/10/orange-beagle/
238        http://mashable.com/2013/01/10/people-winter/
253    http://mashable.com/2013/01/10/why-men-have-ni...
261    http://mashable.com/2013/01/11/a-cappella-remi...
264    http://mashable.com/2013/01/11/basketball-buzz...
292         http://mashable.com/2013/01/11/puppy-stairs/
304    http://mashable.com/2013/01/11/tim-hortons-sin...
310    http://mashable.com/2013/01/11/youtube-office-...
342    http://mashable.com/2013/01/13/joke-cubicle-ma...
400        http://mashable.com/2013/01/14/kid-bulldozer/
Name: url, dtype: object

For images :
 4      http://mashable.com/2013/01/07/att-u-verse-apps/
6     http://mashable.com/2013/01/07/bodymedia-armba...
7      http://mashable.com/2013/01/07/canon-poweshot-n/
12        http://mashable.com/2013/01/07/creature-cups/
20         http://mashable.com/2013/01/07/ford-glympse/
22    http://mashable.com/2013/01/07/fujifilm-50x-su...
27    http://mashable.com/2013/01/07/intel-awesome-l...
31    http://mashable.com/2013/01/07/lego-taking-rob...
32     http://mashable.com/2013/01/07/lg-oled-tv-price/
40    http://mashable.com/2013/01/07/monster-ea-gami...
42    http://mashable.com/2013/01/07/nectar-mobile-p...
43    http://mashable.com/2013/01/07/nvidia-project-...
44       http://mashable.com/2013/01/07/nvidia-tegra-4/
46    http://mashable.com/2013/01/07/omron-fitness-m...
48    http://mashable.com/2013/01/07/panasonic-secon...
Name: url, dtype: object


data[data['num_imgs']==0]['url'][:5]

5      http://mashable.com/2013/01/07/beewi-smart-toys/
8     http://mashable.com/2013/01/07/car-of-the-futu...
28    http://mashable.com/2013/01/07/isp02-iphone-ta...
37    http://mashable.com/2013/01/07/man-instagram-i...
39    http://mashable.com/2013/01/07/messi-year-all-...
Name: url, dtype: object


data['num_imgs'] = data['num_imgs'].replace(0, 1)


print("Videos :\n",data['num_videos'].describe(),"\n90%\t   ",data['num_videos'].quantile(0.9),"\n")
print("Images :\n",data['num_imgs'].describe(),"\n90%\t   ",data['num_imgs'].quantile(0.9))

Videos :
 count    35103.000000
mean         1.187762
std          3.962538
min          0.000000
25%          0.000000
50%          0.000000
75%          1.000000
max         75.000000
Name: num_videos, dtype: float64 
90%	    2.0 

Images :
 count    35103.000000
mean         4.505114
std          8.009084
min          1.000000
25%          1.000000
50%          1.000000
75%          3.000000
max        128.000000
Name: num_imgs, dtype: float64 
90%	    13.0


pd.set_option('display.max_colwidth', 100)
data[data['num_videos']==3]['url'][:5]

195                   http://mashable.com/2013/01/10/chilly-gonzales-ipad/
222     http://mashable.com/2013/01/10/kickstart-films-oscars-nominations/
877           http://mashable.com/2013/01/22/the-second-space-race-its-on/
990            http://mashable.com/2013/01/24/faa-stunt-pilot-viral-video/
1230        http://mashable.com/2013/01/28/star-trek-into-darkness-iphone/
Name: url, dtype: object


pd.set_option('display.max_colwidth', 100)
data[data['num_videos']==1]['url'][0:10]

9                   http://mashable.com/2013/01/07/chuck-hagel-website/
21                     http://mashable.com/2013/01/07/ftc-google-leaks/
30                http://mashable.com/2013/01/07/land-a-job-at-spotify/
37           http://mashable.com/2013/01/07/man-instagram-images-blind/
39                 http://mashable.com/2013/01/07/messi-year-all-goals/
50                             http://mashable.com/2013/01/07/papertab/
59    http://mashable.com/2013/01/07/samsung-2013-ces-press-conference/
66     http://mashable.com/2013/01/07/sharp-unveils-next-gen-igo-hdtvs/
76               http://mashable.com/2013/01/07/vh1-new-logo-plus-sign/
77                    http://mashable.com/2013/01/07/youtube-future-tv/
Name: url, dtype: object


colonnesIQR=['num_videos','num_imgs']
IQRparameters = IQRrule(colonnesIQR)
IQR_vids = IQRparameters['num_videos']
IQR_imgs = IQRparameters['num_imgs']

For 'num_videos' :
Lower limit : -2
Upper limit : 3
The upper quantile for 'upper_lim' : 94.2027746916218 %
The upper quantile for 'lower_lim' : 0.0 %

For 'num_imgs' :
Lower limit : -2
Upper limit : 6
The upper quantile for 'upper_lim' : 79.62282426003475 %
The upper quantile for 'lower_lim' : 0.0 %


row_vid_count = data[data['num_videos']>IQR_vids[0]].shape[0]
print(f"Number of rows with {IQR_vids[0]+1} or more videos :",row_vid_count)
print("Which represents",round(row_vid_count*100/len(data),2),"% of the dataset.\n")
row_img_count = data[data['num_imgs']>IQR_imgs[0]].shape[0]
print(f"Number of rows with {IQR_imgs[0]+1} or more images :",row_img_count)
print("Which represents",round(row_img_count*100/len(data),2),"% of the dataset.")

fig, axes = plt.subplots(2,2, figsize=(16,9))
sns.boxplot(x=data['num_videos'],ax=axes[0, 0])
axes[0, 0].set_title("Distribution of 'num_videos' column")
sns.boxplot(x=data['num_videos'],ax=axes[0, 1])
axes[0, 1].set_xlim(0,data['num_videos'].quantile(1-IQR_vids[2]))
axes[0, 1].set_title(f"Distribution of 'num_videos' to the upper quantile from IQR Rule ({round(100-(IQR_vids[2]*100),2)}%)")

sns.boxplot(x=data['num_imgs'],ax=axes[1, 0])
axes[1, 0].set_title("Distribution of 'num_imgs' column")
sns.boxplot(x=data['num_imgs'],ax=axes[1, 1])
axes[1, 1].set_xlim(0,data['num_imgs'].quantile(1-IQR_imgs[2]))
axes[1, 1].set_title(f"Distribution of 'num_imgs' to the upper quantile from IQR Rule ({round(100-(IQR_imgs[2]*100),2)}%)")

plt.tight_layout()
plt.show()

Number of rows with 4 or more videos : 2035
Which represents 5.8 % of the dataset.

Number of rows with 7 or more images : 7153
Which represents 20.38 % of the dataset.


commun_bad_rows = data[(data['num_videos']>IQR_vids[0]) & (data['num_imgs']>IQR_imgs[0])].shape[0]
print(f"Number of rows with {IQR_vids[0]+1} or more videos, and {IQR_imgs[0]+1} or more images :",commun_bad_rows)
print("Which represents",round(commun_bad_rows*100/len(data),2),"% of the dataset.\n")

Number of rows with 4 or more videos, and 7 or more images : 197
Which represents 0.56 % of the dataset.


print("Mean of the 'shares' for articles with 7 or more images :",data[data['num_imgs'] >6]['shares'].mean())
print("Mean of the 'shares' for articles with 6 images max :",data[data['num_imgs'] <=6]['shares'].mean(),"\n")
print("Mean of the 'shares' for articles with 4 or more videos :",data[data['num_videos'] >3]['shares'].mean())
print("Mean of the 'shares' for articles with 3 videos max :",data[data['num_videos'] <=3]['shares'].mean(),"\n")
print("Mean of the 'shares' :",data['shares'].mean(),"\n-----------------------------------------------------------------------------")

print("Median of the 'shares' for articles with 7 or more images :",data[data['num_imgs'] >6]['shares'].median())
print("Median of the 'shares' for articles with 6 images max :",data[data['num_imgs'] <=6]['shares'].median(),"\n")
print("Median of the 'shares' for articles with 4 or more videos :",data[data['num_videos'] >3]['shares'].median())
print("Median of the 'shares' for articles with 3 videos max :",data[data['num_videos'] <=3]['shares'].median(),"\n")
print("Median of the 'shares' :",data['shares'].median())

Mean of the 'shares' for articles with 7 or more images : 1840.925066405704
Mean of the 'shares' for articles with 6 images max : 1628.7341323792486 

Mean of the 'shares' for articles with 4 or more videos : 1692.002457002457
Mean of the 'shares' for articles with 3 videos max : 1670.740020563687 

Mean of the 'shares' : 1671.9726519100932 
-----------------------------------------------------------------------------
Median of the 'shares' for articles with 7 or more images : 1500.0
Median of the 'shares' for articles with 6 images max : 1300.0 

Median of the 'shares' for articles with 4 or more videos : 1300.0
Median of the 'shares' for articles with 3 videos max : 1300.0 

Median of the 'shares' : 1300.0


def nb_imgs_url(url):
    response = requests.get(url)
    html = response.content
    soup = BeautifulSoup(html, 'html.parser')
    img_tags = soup.find_all('img')
    num_images = len(img_tags) - 2
    return num_images

def nb_videos_url(url):
    response = requests.get(url)
    html = response.content
    text=str(html)
    return text.count("video-container")/2


#df_imgs_videos_test = data[(data['num_imgs']>7)][0:1000].copy()
#df_imgs_videos_test["true_nb_imgs"] = df_imgs_videos_test['url'].apply(lambda url: nb_imgs_url(url))

# It takes approximately 15 min to run


#df_imgs_videos_test = data[(data['num_videos']>4)][0:1000].copy()
#df_imgs_videos_test["true_nb_videos"] = df_imgs_videos_test['url'].apply(lambda url: nb_videos_url(url))
#df_imgs_videos_test["true_nb_videos"].describe()
# It takes approximately 15 min to run


DropColumn(data, ['url'])


data_no_theme = data[(data['data_channel_is_bus']==0) & (data['data_channel_is_world']==0) & (data['data_channel_is_socmed']==0) & (data['data_channel_is_entertainment']==0) & (data['data_channel_is_tech']==0) & (data['data_channel_is_lifestyle']==0)]
print("Number of articles with no theme mentionned :",data_no_theme.shape[0])
data_no_LDA = data[(data['LDA_00']==0) & (data['LDA_00']==0) & (data['LDA_01']==0) & (data['LDA_02']==0) & (data['LDA_03']==0) & (data['LDA_04']==0)]
print("Number of articles with no LDA mentionned :",data_no_LDA.shape[0])

Number of articles with no theme mentionned : 4793
Number of articles with no LDA mentionned : 0


cols_channel_LDA = ['data_channel_is_lifestyle',
       'data_channel_is_entertainment', 'data_channel_is_bus',
       'data_channel_is_socmed', 'data_channel_is_tech',
       'data_channel_is_world', 'LDA_00',
       'LDA_01', 'LDA_02', 'LDA_03', 'LDA_04']
df_channel_LDA = data[cols_channel_LDA]
print(df_channel_LDA.shape)
df_channel_LDA.head()

(35103, 11)


corr = df_channel_LDA.corr()

plt.figure(figsize=(20, 6))
mask = np.triu(np.ones_like(corr, dtype=bool))
ax = sns.heatmap(corr, annot=True, fmt='.2f', cmap='coolwarm', cbar=True, mask=mask)
ax.set_title('Correlation Heatmap')
plt.show()


lda_columns = ["LDA_00", "LDA_01", "LDA_02", "LDA_03", "LDA_04"]
data_channel_columns = ['data_channel_is_lifestyle', 'data_channel_is_entertainment', 'data_channel_is_bus',
                  'data_channel_is_socmed', 'data_channel_is_tech', 'data_channel_is_world']


for lda_column in lda_columns:
    plt.figure(figsize=(12, 6))
    sns.boxplot(x=lda_column, y="variable", hue="value", data=pd.melt(df_channel_LDA, id_vars=[lda_column], value_vars=data_channel_columns))
    plt.title(f"Boxplot of {lda_column} across Binary Categories")
    plt.show()


for lda_column in lda_columns:
    plt.figure(figsize=(12, 6))
    for data_channel in data_channel_columns:
        sns.kdeplot(df_channel_LDA[df_channel_LDA[data_channel] == 1][lda_column], label=data_channel)
    plt.title(f"Density Plot of {lda_column}")
    plt.legend()
    plt.show()


DropColumn(data, lda_columns)


print(data.shape)
data.head()

(35103, 53)


data["theme"] = "Other"

for index, col in enumerate(data_channel_columns):
    for i in range(len(data)):
        if data[col].iloc[i]==1:
            data["theme"].iloc[i]=col[16::].capitalize()

A value is trying to be set on a copy of a slice from a DataFrame

See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy


filter_data_channel_theme = ['data_channel_is_lifestyle', 'data_channel_is_entertainment', 'data_channel_is_bus',
'data_channel_is_socmed', 'data_channel_is_tech', 'data_channel_is_world', 'theme']
data[filter_data_channel_theme].head(10)


DropColumn(data,data_channel_columns)
print(data.shape)
data.head()

(35103, 48)


fig, (ax1,ax2) = plt.subplots(1,2, figsize=(22,8))

num_themes = data['theme'].nunique()
colors_gradient = [plt.cm.coolwarm(i / float(num_themes - 1)) for i in range(num_themes)]
theme_group = data.groupby('theme')['shares'].mean().sort_values(ascending=True)
theme_group.plot(kind='barh', color=colors_gradient, ax=ax1)
ax1.set_xlabel('Average Popularity')
ax1.set_title('Average Popularity by Articles Themes')

theme_counts = data['theme'].value_counts().sort_index()
ax2.pie(theme_counts, labels=['No theme', 'Lifestyle', 'Entertainment', 'Bus', 'Socmed', 'Tech', 'World'],
        autopct=lambda p: '{:.0f}%  ({:,.0f})'.format(p, p * sum(theme_counts) / 100),
        startangle=30, colors=plt.cm.Spectral(np.linspace(0, 1, data['theme'].nunique())))
ax2.set_title('Pie chart of Articles Distribution by Articles Themes')

plt.show()


keyword_columns = ['kw_min_min', 'kw_max_min', 'kw_avg_min',
                    'kw_min_max', 'kw_max_max', 'kw_avg_max',
                    'kw_min_avg', 'kw_max_avg', 'kw_avg_avg',
                    'num_keywords']
data[keyword_columns].describe()


plt.figure(figsize=(16,6))
sns.heatmap(data[keyword_columns].corr(),cmap="coolwarm",annot=True)
plt.show()


cols_kw_to_drop=['kw_min_min','kw_max_min','kw_min_max','kw_max_max']
DropColumn(data,cols_kw_to_drop)


keyword_columns = list(set(keyword_columns) - set(cols_kw_to_drop))
IQRparameters = IQRrule(keyword_columns)

For 'kw_avg_min' :
Lower limit : -181
Upper limit : 677
The upper quantile for 'upper_lim' : 94.87223314246646 %
The upper quantile for 'lower_lim' : 0.0 %

For 'kw_min_avg' :
Lower limit : -2988
Upper limit : 4980
The upper quantile for 'upper_lim' : 100.0 %
The upper quantile for 'lower_lim' : 0.0 %

For 'num_keywords' :
Lower limit : 1
Upper limit : 14
The upper quantile for 'upper_lim' : 100.0 %
The upper quantile for 'lower_lim' : 0.0 %

For 'kw_avg_max' :
Lower limit : -61654
Upper limit : 560932
The upper quantile for 'upper_lim' : 97.51588183346152 %
The upper quantile for 'lower_lim' : 0.0 %

For 'kw_max_avg' :
Lower limit : 48
Upper limit : 9334
The upper quantile for 'upper_lim' : 94.07173176081817 %
The upper quantile for 'lower_lim' : 0.21080819303193457 %

For 'kw_avg_avg' :
Lower limit : 644
Upper limit : 5198
The upper quantile for 'upper_lim' : 95.85790388285902 %
The upper quantile for 'lower_lim' : 0.2307495085890095 %


print("Upper limit for 'shares' :",upper_lim_shares)
rows_count_avgmax_beyond_upperlimshares = data[data['kw_avg_max']>upper_lim_shares].shape[0]
print("Number of rows where 'kw_avg_max' is beyond the upper limit of 'shares' :",rows_count_avgmax_beyond_upperlimshares)
print("Which represent",100*rows_count_avgmax_beyond_upperlimshares/len(data),"% of the dataset.\n")
rows_count_maxavg_beyond_upperlimshares = data[data['kw_max_avg']>upper_lim_shares].shape[0]
print("Number of rows where 'kw_max_avg' is beyond the upper limit of 'shares' :",rows_count_maxavg_beyond_upperlimshares)
print("Which represent",100*rows_count_maxavg_beyond_upperlimshares/len(data),"% of the dataset.\n")

Upper limit for 'shares' : 5581.0
Number of rows where 'kw_avg_max' is beyond the upper limit of 'shares' : 34980
Which represent 99.64960259806854 % of the dataset.

Number of rows where 'kw_max_avg' is beyond the upper limit of 'shares' : 9992
Which represent 28.464803578041764 % of the dataset.


DropColumn(data, ['kw_avg_max', 'kw_avg_min'])
keyword_columns.remove('kw_avg_max')
keyword_columns.remove('kw_avg_min')


data['kw_min_avg'].replace(-1, 0, inplace=True)


reference_columns =  ['num_hrefs','num_self_hrefs','self_reference_min_shares','self_reference_max_shares','self_reference_avg_sharess']


data[reference_columns].describe()


for col in reference_columns:
    fig, (ax1,ax2) = plt.subplots(1,2,figsize=(20,6))
    ax1.scatter(data[col],data["shares"])
    ax2.hist(data[col],bins='auto')
    ax1.set_xlabel(col)
    ax1.set_ylabel("Shares")
    ax1.set_title(f"Shares compared to {col}")
    ax2.set_xlabel(col)
    ax2.set_ylabel("Frequency")
    ax2.set_title(f"Distribution of {col}")
    plt.show()


corr_self_references = data[reference_columns].corr()

plt.figure(figsize=(20, 6))
mask = np.triu(np.ones_like(corr_self_references, dtype=bool))
ax = sns.heatmap(corr_self_references, annot=True, fmt='.2f', cmap='coolwarm', cbar=True, mask=mask)

ax.set_title('Correlation Heatmap')

plt.show()


DropColumn(data,['self_reference_max_shares','self_reference_min_shares'])
reference_columns.remove('self_reference_max_shares')
reference_columns.remove('self_reference_min_shares')


IQRparameters = IQRrule(reference_columns)
IQR_num_hrefs = IQRparameters['num_hrefs']
IQR_num_self_href = IQRparameters['num_self_hrefs']
IQR_refs_avg = IQRparameters['self_reference_avg_sharess']

For 'num_hrefs' :
Lower limit : -10
Upper limit : 27
The upper quantile for 'upper_lim' : 93.8039483804803 %
The upper quantile for 'lower_lim' : 0.0 %

For 'num_self_hrefs' :
Lower limit : -4
Upper limit : 9
The upper quantile for 'upper_lim' : 95.94336666381791 %
The upper quantile for 'lower_lim' : 0.0 %

For 'self_reference_avg_sharess' :
Lower limit : -4750
Upper limit : 10457
The upper quantile for 'upper_lim' : 89.53935561063157 %
The upper quantile for 'lower_lim' : 0.0 %


weekday_columns = ['weekday_is_monday', 'weekday_is_tuesday', 'weekday_is_wednesday',
       'weekday_is_thursday', 'weekday_is_friday', 'weekday_is_saturday',
       'weekday_is_sunday']

for weekday in weekday_columns:
    print(f"{weekday} :",data[weekday].unique(), "unique values.")

weekday_is_monday : [1. 0.] unique values.
weekday_is_tuesday : [0. 1.] unique values.
weekday_is_wednesday : [0. 1.] unique values.
weekday_is_thursday : [0. 1.] unique values.
weekday_is_friday : [0. 1.] unique values.
weekday_is_saturday : [0. 1.] unique values.
weekday_is_sunday : [0. 1.] unique values.


count=0
for col in weekday_columns:
  count+=data[col].value_counts()[1]
if count == data.shape[0]: print("Every article has a signle day")
else: print("Some articles are based on multiple days.")

Every article has a signle day


data["weekday"] = 0

for index, col in enumerate(weekday_columns):
    for i in range(len(data)):
        if data[col].iloc[i]==1:
            data["weekday"].iloc[i]=index+1

A value is trying to be set on a copy of a slice from a DataFrame

See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy


filter_weekdays = ['weekday_is_monday', 'weekday_is_tuesday', 'weekday_is_wednesday',
       'weekday_is_thursday', 'weekday_is_friday', 'weekday_is_saturday',
       'weekday_is_sunday', "weekday"]
for weekday in filter_weekdays:
    print(f"{weekday} :\n",data[weekday].value_counts())

weekday_is_monday :
 0.0    29239
1.0     5864
Name: weekday_is_monday, dtype: int64
weekday_is_tuesday :
 0.0    28527
1.0     6576
Name: weekday_is_tuesday, dtype: int64
weekday_is_wednesday :
 0.0    28466
1.0     6637
Name: weekday_is_wednesday, dtype: int64
weekday_is_thursday :
 0.0    28602
1.0     6501
Name: weekday_is_thursday, dtype: int64
weekday_is_friday :
 0.0    30027
1.0     5076
Name: weekday_is_friday, dtype: int64
weekday_is_saturday :
 0.0    32990
1.0     2113
Name: weekday_is_saturday, dtype: int64
weekday_is_sunday :
 0.0    32767
1.0     2336
Name: weekday_is_sunday, dtype: int64
weekday :
 3    6637
2    6576
4    6501
1    5864
5    5076
7    2336
6    2113
Name: weekday, dtype: int64


weekdays_to_drop = ['weekday_is_monday', 'weekday_is_tuesday', 'weekday_is_wednesday',
       'weekday_is_thursday', 'weekday_is_friday', 'weekday_is_saturday',
       'weekday_is_sunday', "is_weekend"]
data.drop(weekdays_to_drop, axis=1, inplace=True)
print(data.shape)
data.head()

(35103, 33)


grouped_data = data.groupby('weekday').agg({'shares': 'mean'}).rename(columns={'shares': 'avg_shares'})
grouped_data['num_articles'] = data['weekday'].value_counts().sort_index()

plt.figure(figsize=(16, 6))
ax1 = grouped_data['num_articles'].plot(kind='line', color='blue', grid=True, label='Number of Articles')
ax2 = grouped_data['avg_shares'].plot(kind='line', color='orange', grid=True, secondary_y=True, label='Average Shares')

ax1.set_xlabel('Day of Week')
ax1.set_ylabel('Number of Articles')
ax2.set_ylabel('Average Shares')

plt.title('Average Shares vs Number of articles, by the Weekday')
ax1.legend(loc='upper left')
ax2.legend(loc='upper right')

plt.show()


weekday_counts = data['weekday'].value_counts().sort_index()

plt.figure(figsize=(8, 8))
plt.pie(weekday_counts, labels=['Monday', 'Tuesday', 'Wednesday', 'Thursday', 'Friday', 'Saturday', 'Sunday'],
        autopct=lambda p: '{:.0f}%  ({:,.0f})'.format(p, p * sum(weekday_counts) / 100),
        startangle=30)

plt.title('Pie chart of Articles Distribution by Weekday')
plt.show()


rate_columns = ['global_rate_positive_words','global_rate_negative_words','rate_positive_words','rate_negative_words']


data[rate_columns].describe()


for col in rate_columns:
    fig, (ax1,ax2) = plt.subplots(1,2,figsize=(20,6))
    ax1.scatter(data[col],data["shares"])
    ax2.hist(data[col],bins='auto')
    ax1.set_xlabel(col)
    ax1.set_ylabel("Shares")
    ax1.set_title(f"Shares compared to {col}")
    ax2.set_xlabel(col)
    ax2.set_ylabel("Frequency")
    ax2.set_title(f"Distribution of {col}")
    plt.show()


corr_rate = data[rate_columns].corr()

plt.figure(figsize=(20, 6))
mask = np.triu(np.ones_like(corr_rate, dtype=bool))
ax = sns.heatmap(corr_rate, annot=True, fmt='.2f', cmap='coolwarm', cbar=True, mask=mask)

ax.set_title('Correlation Heatmap')

plt.show()


DropColumn(data,['rate_positive_words','rate_negative_words'])
rate_columns = list(set(rate_columns) - set(['rate_positive_words','rate_negative_words']))


polarity_columns = ['global_sentiment_polarity','avg_positive_polarity','min_positive_polarity','max_positive_polarity','avg_negative_polarity', 'min_negative_polarity','max_negative_polarity','title_sentiment_polarity', 'abs_title_sentiment_polarity']


data[polarity_columns].describe()


for col in polarity_columns:
    fig, (ax1,ax2) = plt.subplots(1,2,figsize=(20,6))
    ax1.scatter(data[col],data["shares"])
    ax2.hist(data[col],bins='auto')
    ax1.set_xlabel(col)
    ax1.set_ylabel("Shares")
    ax1.set_title(f"Shares compared to {col}")
    ax2.set_xlabel(col)
    ax2.set_ylabel("Frequency")
    ax2.set_title(f"Distribution of {col}")
    plt.show()


corr_polarity = data[polarity_columns].corr()

plt.figure(figsize=(20, 6))
mask = np.triu(np.ones_like(corr_polarity, dtype=bool))
ax = sns.heatmap(corr_polarity, annot=True, fmt='.2f', cmap='coolwarm', cbar=True, mask=mask)

ax.set_title('Correlation Heatmap')

plt.show()


polarities_to_delete=["min_positive_polarity","max_positive_polarity","min_negative_polarity","max_negative_polarity","abs_title_sentiment_polarity"]
DropColumn(data,polarities_to_delete)


subjectivity_columns = ['global_subjectivity','title_subjectivity','abs_title_subjectivity']


data[subjectivity_columns].describe()


for col in subjectivity_columns:
    fig, (ax1,ax2) = plt.subplots(1,2,figsize=(20,6))
    ax1.scatter(data[col],data["shares"])
    ax2.hist(data[col],bins='auto')
    ax1.set_xlabel(col)
    ax1.set_ylabel("Shares")
    ax1.set_title(f"Shares compared to {col}")
    ax2.set_xlabel(col)
    ax2.set_ylabel("Frequency")
    ax2.set_title(f"Distribution of {col}")
    plt.show()


plt.figure(figsize=(16,6))
sns.heatmap(data[subjectivity_columns].corr(),cmap="coolwarm",annot=True)
plt.title('Correlation Heatmap')
plt.show()


DropColumn(data,["abs_title_subjectivity"])


print(data.shape)

(35103, 25)


data_capped = data.copy() #Cap the outliers to the IQR's limits
data_quality = data.copy() #Deleting the outliers out of the IQR's limits


#TOKENS
data_capped.loc[data_capped['n_tokens_title'] > IQR_n_tokens_title[0], 'n_tokens_title'] = IQR_n_tokens_title[0]
data_capped.loc[data_capped['n_tokens_title'] < IQR_n_tokens_title[1], 'n_tokens_title'] = IQR_n_tokens_title[1]
data_capped.loc[data_capped['n_tokens_content'] > IQR_n_tokens_content[0], 'n_tokens_content'] = IQR_n_tokens_content[0]
data_capped.loc[data_capped['average_token_length'] > IQR_avg_token_length[0], 'average_token_length'] = IQR_avg_token_length[0]
data_capped.loc[data_capped['average_token_length'] < IQR_avg_token_length[1], 'average_token_length'] = IQR_avg_token_length[1]

data_quality = data_quality[data_quality['n_tokens_title']<=IQR_n_tokens_title[0]]
data_quality = data_quality[data_quality['n_tokens_title']>=IQR_n_tokens_title[1]]
data_quality = data_quality[data_quality['n_tokens_content']<=IQR_n_tokens_content[0]]
data_quality = data_quality[data_quality['average_token_length']<=IQR_avg_token_length[0]]
data_quality = data_quality[data_quality['average_token_length']>=IQR_avg_token_length[1]]

#IMAGES & VIDS (here we use our results from the webscrapping!)
data_capped.loc[data_capped['num_videos'] > IQR_vids[0], 'num_videos'] = 2 #The mean of the webscrapping
data_capped.loc[data_capped['num_imgs'] > IQR_imgs[0], 'num_imgs'] = 0 #The mean of the webscrapping

data_quality = data_quality[data_quality ['num_videos']<=IQR_vids[0]]
data_quality = data_quality[data_quality['num_imgs']<=IQR_imgs[0]]

#KEYWORDS
data_capped.loc[data_capped['kw_avg_avg'] > upper_lim_shares, 'kw_avg_avg'] = upper_lim_shares
data_capped.loc[data_capped['kw_avg_avg'] < lower_lim_shares, 'kw_avg_avg'] = lower_lim_shares
data_capped.loc[data_capped['kw_max_avg'] > upper_lim_shares, 'kw_max_avg'] = upper_lim_shares
data_capped.loc[data_capped['kw_max_avg'] < lower_lim_shares, 'kw_max_avg'] = lower_lim_shares

data_quality = data_quality[data_quality['kw_avg_avg']<=upper_lim_shares]
data_quality = data_quality[data_quality['kw_max_avg']<=upper_lim_shares]
data_quality = data_quality[data_quality['kw_avg_avg']>=lower_lim_shares]
data_quality = data_quality[data_quality['kw_max_avg']>=lower_lim_shares]

#REFERENCES
data_capped.loc[data_capped["self_reference_avg_sharess"] > IQR_refs_avg[0], "self_reference_avg_sharess"] = IQR_refs_avg[0]
data_capped.loc[data_capped["num_self_hrefs"] > IQR_num_self_href[0], "num_self_hrefs"] = IQR_num_self_href[0]
data_capped.loc[data_capped["num_hrefs"] > IQR_num_hrefs[0], "num_hrefs"] = IQR_num_hrefs[0]

data_quality = data_quality[data_quality['self_reference_avg_sharess']<=IQR_refs_avg[0]]
data_quality = data_quality[data_quality['num_self_hrefs']<=IQR_num_self_href[0]]
data_quality = data_quality[data_quality['num_hrefs']<=IQR_num_hrefs[0]]

print("'data_quality' shape:",data_quality.shape,"\nIt represents",round(100*data_quality.shape[0]/len(data),2),"% compared to the dataset without 'shares' outliers only.")

'data_quality' shape: (15111, 25) 
It represents 43.05 % compared to the dataset without 'shares' outliers only.


df = data_quality.copy()  #for the data visulisations after


fig, (ax1,ax2) = plt.subplots(1,2,figsize=(25, 15))

heatmap_capped = sns.heatmap(data_capped.corr()[['shares']].sort_values(by='shares', ascending=False), vmin=-1, vmax=1, annot=True, cbar=None, ax=ax1)
heatmap_capped.set_title("'data_capped' Features Correlating with Shares", fontdict={'fontsize':14}, pad=16);

heatmap_quality = sns.heatmap(data_quality.corr()[['shares']].sort_values(by='shares', ascending=False), vmin=-1, vmax=1, annot=True, ax=ax2)
heatmap_quality.set_title("'data_quality' Features Correlating with Shares", fontdict={'fontsize':14}, pad=16);

plt.tight_layout()
plt.show()

The default value of numeric_only in DataFrame.corr is deprecated. In a future version, it will default to False. Select only valid columns or specify the value of numeric_only to silence this warning.
The default value of numeric_only in DataFrame.corr is deprecated. In a future version, it will default to False. Select only valid columns or specify the value of numeric_only to silence this warning.


plt.figure(figsize=(25,10))
sns.heatmap(data_quality.corr(), cmap='RdBu_r', annot=True)
plt.title("Correlation Heatmap of 'data_quality'")
plt.show()

The default value of numeric_only in DataFrame.corr is deprecated. In a future version, it will default to False. Select only valid columns or specify the value of numeric_only to silence this warning.


plt.figure(figsize=(25,10))
sns.heatmap(data_capped.corr(),cmap='RdBu_r',annot=True)
plt.title("Correlation Heatmap of 'data_quality'")
plt.show()

The default value of numeric_only in DataFrame.corr is deprecated. In a future version, it will default to False. Select only valid columns or specify the value of numeric_only to silence this warning.


array_to_boxplot=[['shares','kw_avg_avg','kw_max_avg','kw_min_avg'],
['n_tokens_content'],
['average_token_length','n_tokens_title'],
['n_unique_tokens'],
['num_videos', 'num_imgs'],
['self_reference_avg_sharess'],
['num_self_hrefs','num_hrefs'],
['global_rate_positive_words','global_rate_negative_words'],
['global_sentiment_polarity','avg_positive_polarity','avg_negative_polarity','title_sentiment_polarity'],
['global_subjectivity','title_subjectivity']]

for category in array_to_boxplot:
    fig, (ax1,ax2) = plt.subplots(1,2,figsize=(20,4))
    sns.boxplot(data_quality[category],ax=ax1)
    sns.boxplot(data_capped[category],ax=ax2)
    ax1.set_ylabel("Distribution")
    ax1.set_title(f"Boxplot of {category} in 'data_quality'")
    ax2.set_ylabel("Distribution")
    ax2.set_title(f"Boxplot of {category} in 'data_capped'")
    plt.tight_layout()
    plt.show()


le = LabelEncoder()
data_quality['theme'] = le.fit_transform(data_quality['theme'])
data_capped['theme'] = le.fit_transform(data_capped['theme'])

normalized_data_quality = data_quality.copy()
normalized_data_capped = data_capped.copy()

sc = StandardScaler()
normalized_data_quality = sc.fit_transform(normalized_data_quality)
normalized_data_capped = sc.fit_transform(normalized_data_capped)

fig, (ax1,ax2) = plt.subplots(2,1,figsize=(25,10))
sns.boxplot(normalized_data_quality,ax=ax1)
sns.boxplot(normalized_data_capped,ax=ax2)
ax1.set_ylabel("Distribution")
ax1.set_title(f"Boxplot of normalized features in 'data_quality'")
ax2.set_ylabel("Distribution")
ax2.set_title(f"Boxplot of normalized features in 'data_capped'")
ax1.set_xticklabels(data_quality.columns, rotation=45, ha='right')
ax2.set_xticklabels(data_capped.columns, rotation=45, ha='right')
plt.tight_layout()
plt.show()


def SplitAndScaler(df, testSize, classifierModel):
    y = df['shares']
    X= df.drop('shares', axis=1)
    if classifierModel == True:
      X_train, X_test, y_train, y_test = train_test_split(X, y,test_size=testSize ,random_state =42, stratify=y)
      #'stratify=y' maintains the same proportion of each class of 'shares' in the test and train sets.
    else :
      X_train, X_test, y_train, y_test = train_test_split(X, y,test_size=testSize ,random_state =42)

    sc = StandardScaler()
    X_train = sc.fit_transform(X_train)
    X_test = sc.transform(X_test)
    return X_train, X_test, y_train, y_test


X_train, X_test, y_train, y_test = SplitAndScaler(data_quality, 0.3, False)


lr_model_quality = LinearRegression()
lr_model_quality.fit(X_train, y_train)

cv_lr_quality = cross_val_score(estimator = lr_model_quality, X = X_train, y = y_train, cv = 6)

y_pred_lr_quality_train = lr_model_quality.predict(X_train)
r2_score_lr_quality_train = r2_score(y_train, y_pred_lr_quality_train)

y_pred_lr_quality_test = lr_model_quality.predict(X_test)
r2_score_lr_quality_test = r2_score(y_test, y_pred_lr_quality_test)

rmse_lr_quality = (np.sqrt(mean_squared_error(y_test, y_pred_lr_quality_test)))
print("CV: ", cv_lr_quality.mean())
print('R2_score (train): ', r2_score_lr_quality_train)
print('R2_score (test): ', r2_score_lr_quality_test)
print("RMSE: ", rmse_lr_quality)

CV:  0.0651662169400634
R2_score (train):  0.07163107420126569
R2_score (test):  0.07992582117395641
RMSE:  967.4855132806035


lasso_pipe_quality = Lasso(random_state = 42, max_iter=3000)
lasso_pipe_quality.fit(X_train, y_train)

cv_lasso_quality = cross_val_score(estimator = lasso_pipe_quality, X = X_train, y = y_train, cv = 6)

y_pred_lasso_quality_train = lasso_pipe_quality.predict(X_train)
r2_score_lasso_quality_train = r2_score(y_train, y_pred_lasso_quality_train)

y_pred_lasso_quality_test = lasso_pipe_quality.predict(X_test)
r2_score_lasso_quality_test = r2_score(y_test, y_pred_lasso_quality_test)

rmse_lasso_quality = (np.sqrt(mean_squared_error(y_test, y_pred_lasso_quality_test)))
print('CV: ', cv_lasso_quality.mean())
print('R2_score (train): ', r2_score_lasso_quality_train)
print('R2_score (test): ', r2_score_lasso_quality_test)
print("RMSE: ", rmse_lasso_quality)

CV:  0.06532596076230629
R2_score (train):  0.07158671947429718
R2_score (test):  0.07983313321346641
RMSE:  967.5342441344675


dt_quality = DecisionTreeRegressor(random_state = 42)
dt_quality.fit(X_train, y_train)

cv_dt_quality = cross_val_score(estimator = dt_quality, X = X_train, y = y_train, cv = 6)

y_pred_dt_quality_train = dt_quality.predict(X_train)
r2_score_dt_quality_train = r2_score(y_train, y_pred_dt_quality_train)

y_pred_dt_quality_test = dt_quality.predict(X_test)
r2_score_dt_quality_test = r2_score(y_test, y_pred_dt_quality_test)

rmse_dt_quality = (np.sqrt(mean_squared_error(y_test, y_pred_dt_quality_test)))
print('CV: ', cv_dt_quality.mean())
print('R2_score (train): ', r2_score_dt_quality_train)
print('R2_score (test): ', r2_score_dt_quality_test)
print("RMSE: ", rmse_dt_quality)

CV:  -0.8102815714278028
R2_score (train):  1.0
R2_score (test):  -0.8059407649082078
RMSE:  1355.4540516554207


rf_quality = RandomForestRegressor(random_state = 42)
rf_quality.fit(X_train, y_train)

cv_rf_quality = cross_val_score(estimator = rf_quality,  X = X_train, y = y_train, cv = 6)

y_pred_rf_quality_train = rf_quality.predict(X_train)
r2_score_rf_quality_train = r2_score(y_train, y_pred_rf_quality_train)

y_pred_rf_quality_test = rf_quality.predict(X_test)
r2_score_rf_quality_test = r2_score(y_test, y_pred_rf_quality_test)

rmse_rf_quality = (np.sqrt(mean_squared_error(y_test, y_pred_rf_quality_test)))
print('CV: ', cv_rf_quality.mean())
print('R2_score (train): ', r2_score_rf_quality_train)
print('R2_score (test): ', r2_score_rf_quality_test)
print("RMSE: ", rmse_rf_quality)

CV:  0.09582454966349652
R2_score (train):  0.8758170232396529
R2_score (test):  0.10015705669721842
RMSE:  956.7895140975413


svr_quality = SVR(kernel = 'rbf', gamma = 'scale')
svr_quality.fit(X_train, y_train)

cv_svr_quality = cross_val_score(estimator = svr_quality, X = X_train, y = y_train, cv = 6)

y_pred_svr_quality_train = svr_quality.predict(X_train)
r2_score_svr_quality_train = r2_score(y_train, y_pred_svr_quality_train)

y_pred_svr_quality_test = svr_quality.predict(X_test)
r2_score_svr_quality_test = r2_score(y_test, y_pred_svr_quality_test)

rmse_svr_quality = (np.sqrt(mean_squared_error(y_test, y_pred_svr_quality_test)))
print('CV: ', cv_svr_quality.mean())
print('R2_score (train): ', r2_score_svr_quality_train)
print('R2_score (test): ', r2_score_svr_quality_test)
print("RMSE: ", rmse_svr_quality)

CV:  -0.09777559538790152
R2_score (train):  -0.09285435713221912
R2_score (test):  -0.08015088050758346
RMSE:  1048.2750793447285


modelsReg_quality = [('Linear Regression', rmse_lr_quality, r2_score_lr_quality_train, r2_score_lr_quality_test, cv_lr_quality.mean()),
          ('Lasso Regression', rmse_lasso_quality, r2_score_lasso_quality_train, r2_score_lasso_quality_test, cv_lasso_quality.mean()),
          ('Decision Tree Regression', rmse_dt_quality, r2_score_dt_quality_train, r2_score_dt_quality_test, cv_dt_quality.mean()),
          ('Random Forest Regression', rmse_rf_quality, r2_score_rf_quality_train, r2_score_rf_quality_test, cv_rf_quality.mean()),
          ('Support Vector Regression', rmse_svr_quality, r2_score_svr_quality_train, r2_score_svr_quality_test, cv_svr_quality.mean())
         ]


predictions_reg_quality = pd.DataFrame(data = modelsReg_quality, columns=['Model', 'RMSE', 'R2_Score(training)', 'R2_Score(test)', 'Cross-Validation'])
predictions_reg_quality


plt.figsize=(16,6)
predictions_reg_quality.sort_values(by=['Cross-Validation'], ascending=False, inplace=True)

sns.barplot(x='Cross-Validation', y='Model', data = predictions_reg_quality)
plt.xlabel('Cross-Validaton Score', size=16)
plt.ylabel('Model')
plt.xlim(0,1.0)
plt.show()


f, axes = plt.subplots(2,1, figsize=(16,12))

predictions_reg_quality.sort_values(by=['R2_Score(training)'], ascending=False, inplace=True)

sns.barplot(x='R2_Score(training)', y='Model', data = predictions_reg_quality, palette='Blues_d', ax = axes[0])
axes[0].set_xlabel('R2 Score (Training)', size=16)
axes[0].set_ylabel('Model')
axes[0].set_xlim(0,1.0)

predictions_reg_quality.sort_values(by=['R2_Score(test)'], ascending=False, inplace=True)

sns.barplot(x='R2_Score(test)', y='Model', data = predictions_reg_quality, palette='Reds_d', ax = axes[1])
axes[1].set_xlabel('R2 Score (Test)', size=16)
axes[1].set_ylabel('Model')
axes[1].set_xlim(0,1.0)

plt.show()


predictions_reg_quality.sort_values(by=['RMSE'], ascending=False, inplace=True)
plt.xticks(rotation=45)
sns.barplot(x='Model', y='RMSE', data=predictions_reg_quality)
plt.xlabel('Models', size=16)
plt.ylabel('RMSE', size=16)
plt.show()


df_rf_predictions = pd.DataFrame(rf_quality.predict(X_test[:1000]),list(y_test[:1000]))
df_rf_predictions.reset_index(level=0, inplace=True)
df_rf_predictions = df_rf_predictions.rename(index=str, columns={"index": "Actual shares", 0: "Predicted shares"})
df_rf_predictions.head()


f, ax = plt.subplots(figsize=(17, 3))
sns.regplot(x=df_rf_predictions["Actual shares"], y=df_rf_predictions["Predicted shares"])
plt.show()


plt.figsize=(25,8)
df_rf_predictions.plot(bins=150, kind='hist', alpha=0.8)
plt.title("The 1000th 'Predicted Shares' VS 'Real shares' with the RandomForestRegressor")
plt.xlabel('Number of shares')
plt.show()


feature_importances = rf_quality.feature_importances_
feature_names = data_quality.drop(['shares'], axis=1).columns
sorted_indices = np.argsort(feature_importances)[::-1]  # sort the features by their importance

palette = sns.color_palette("viridis", n_colors=len(feature_importances))
plt.figure(figsize=(16, 8))
plt.title("Feature Importances from Random Forest Regression")
plt.barh(range(len(feature_importances)), feature_importances[sorted_indices], color=palette)
plt.yticks(range(len(feature_importances)), feature_names[sorted_indices])
plt.xlabel('Importance')
plt.ylabel('Features')
plt.gca().invert_yaxis()
plt.show()


new_features_rfr = feature_names[sorted_indices][:15].append(pd.Index(['shares']))
X_train_topfeatures, X_test_topfeatures, y_train_topfeatures, y_test_topfeatures = SplitAndScaler(data_quality[new_features_rfr], 0.3, False)

rfr_top15 = RandomForestRegressor(random_state=42)
rfr_top15.fit(X_train_topfeatures, y_train_topfeatures)

cv_rfr_top15 = cross_val_score(estimator = rfr_top15, X = X_train_topfeatures, y = y_train_topfeatures, cv = 6)

y_pred_rfr_top15_train = rfr_top15.predict(X_train_topfeatures)
r2_score_rfr_top15_train = r2_score(y_train_topfeatures, y_pred_rfr_top15_train)

y_pred_rfr_top15_test = rfr_top15.predict(X_test_topfeatures)
r2_score_rfr_top15_test = r2_score(y_test_topfeatures, y_pred_rfr_top15_test)

rmse_rfr_top15 = (np.sqrt(mean_squared_error(y_test_topfeatures, y_pred_rfr_top15_test)))
print('CV: ', cv_rf_quality.mean())
print('R2_score (train): ', r2_score_rfr_top15_train)
print('R2_score (test): ', r2_score_rfr_top15_test)
print("RMSE: ", rmse_rfr_top15)

CV:  0.09582454966349652
R2_score (train):  0.8720532477670312
R2_score (test):  0.062028696378859105
RMSE:  976.8498750774778


new_features_rfr = list(new_features_rfr.tolist())
new_features_rfr.remove("shares")


explainer_rfr = shap.TreeExplainer(rfr_top15)
shap_values_rfr = explainer_rfr.shap_values(X_test_topfeatures)
X_test_topfeatures_df = pd.DataFrame(X_test_topfeatures, columns=new_features_rfr)
shap.summary_plot(shap_values_rfr, X_test_topfeatures_df)


data_quality_binary_classes = data_quality.copy()


print(data_quality_binary_classes['shares'].mean())
popular=data_quality_binary_classes['shares']>=data_quality_binary_classes['shares'].mean()
unpopular=data_quality_binary_classes['shares']<data_quality_binary_classes['shares'].mean()
data_quality_binary_classes.loc[popular,'shares']=1
data_quality_binary_classes.loc[unpopular,'shares']=0
data_quality_binary_classes['shares'].value_counts()

1540.7552114353782

0    9996
1    5115
Name: shares, dtype: int64


X_train, X_test, y_train, y_test = SplitAndScaler(data_quality_binary_classes, 0.3, True)


def DecisionTreeClassifierModel(X_train, y_train):
    dtclassifier_quality = DecisionTreeClassifier(criterion = 'entropy', random_state = 0)
    dtclassifier_quality.fit(X_train, y_train)

    cv_dt_class_quality = cross_val_score(estimator = dtclassifier_quality, X = X_train, y = y_train, n_jobs=-1, cv=6)

    y_pred_dtclassifier_quality_train = dtclassifier_quality.predict(X_train)
    f1_score_dtclassifier_quality_train = f1_score(y_train, y_pred_dtclassifier_quality_train, average='micro')

    y_pred_dtclassifier_quality_test = dtclassifier_quality.predict(X_test)
    f1_score_dtclassifier_quality_test = f1_score(y_test, y_pred_dtclassifier_quality_test, average='micro')

    accuracy_dtclassifier_quality = accuracy_score(y_test, y_pred_dtclassifier_quality_test)
    print("Accuracy dt_class:", accuracy_dtclassifier_quality)
    print('f1_score dt_class(train): ', f1_score_dtclassifier_quality_train)
    print('f1_score dt_class(test): ', f1_score_dtclassifier_quality_test)
    return accuracy_dtclassifier_quality, f1_score_dtclassifier_quality_train, f1_score_dtclassifier_quality_test, cv_dt_class_quality, y_pred_dtclassifier_quality_test


def KNNClassifierModel(X_train, y_train):
    KNNclassifier_quality = KNeighborsClassifier(n_neighbors = 25)
    KNNclassifier_quality.fit(X_train, y_train)

    cv_knn_class_quality = cross_val_score(estimator = KNNclassifier_quality, X = X_train, y = y_train, n_jobs=-1, cv=6)

    y_pred_knnclassifier_quality_train = KNNclassifier_quality.predict(X_train)
    f1_score_knnclassifier_quality_train = f1_score(y_train, y_pred_knnclassifier_quality_train, average='micro')

    y_pred_knnclassifier_quality_test = KNNclassifier_quality.predict(X_test)
    f1_score_knnclassifier_quality_test = f1_score(y_test, y_pred_knnclassifier_quality_test, average='micro')

    accuracy_knnclassifier_quality = accuracy_score(y_test, y_pred_knnclassifier_quality_test)
    print("Accuracy KNN_class: ", accuracy_knnclassifier_quality)
    print('f1_score KNN_class(train): ', f1_score_knnclassifier_quality_train)
    print('f1_score KNN_class(test): ', f1_score_knnclassifier_quality_test)
    return accuracy_knnclassifier_quality, f1_score_knnclassifier_quality_train, f1_score_knnclassifier_quality_test, cv_knn_class_quality, y_pred_knnclassifier_quality_test


def NaiveBayesClassifierModel(X_train, y_train):
    NBclassifier_quality = GaussianNB()
    NBclassifier_quality.fit(X_train, y_train)

    cv_nb_class_quality = cross_val_score(estimator = NBclassifier_quality, X = X_train, y = y_train, cv = 6)

    y_pred_nbclassifier_quality_train = NBclassifier_quality.predict(X_train)
    f1_score_nbclassifier_quality_train = f1_score(y_train, y_pred_nbclassifier_quality_train, average='micro')

    y_pred_nbclassifier_quality_test = NBclassifier_quality.predict(X_test)
    f1_score_nbclassifier_quality_test = f1_score(y_test, y_pred_nbclassifier_quality_test, average='micro')

    # Metrics results
    accuracy_nbclassifier_quality = accuracy_score(y_test, y_pred_nbclassifier_quality_test)
    print("Accuracy NB_class: ", accuracy_nbclassifier_quality)
    print('f1_score NB_class(train): ', f1_score_nbclassifier_quality_train)
    print('f1_score NB_class(test): ', f1_score_nbclassifier_quality_test)
    return accuracy_nbclassifier_quality, f1_score_nbclassifier_quality_train, f1_score_nbclassifier_quality_test, cv_nb_class_quality, y_pred_nbclassifier_quality_test


def RandomForestClassifierModel(X_train, y_train, param_grid):
    if param_grid== None:
        rfclassifier_quality = RandomForestClassifier(random_state = 42)
    else :
        rfclassifier_quality = RandomForestClassifier(
        n_estimators=param_grid['n_estimators'],
        max_depth=param_grid['max_depth'],
        min_samples_split=param_grid['min_samples_split'],
        min_samples_leaf=param_grid['min_samples_leaf'],
        random_state=param_grid['random_state']
        )

    rfclassifier_quality.fit(X_train, y_train)

    cv_rf_class_quality = cross_val_score(estimator = rfclassifier_quality, X = X_train, y = y_train, cv = 6)

    y_pred_rfclassifier_quality_train = rfclassifier_quality.predict(X_train)
    f1_score_rfclassifier_quality_train = f1_score(y_train, y_pred_rfclassifier_quality_train, average='micro')

    y_pred_rfclassifier_quality_test = rfclassifier_quality.predict(X_test)
    f1_score_rfclassifier_quality_test = f1_score(y_test, y_pred_rfclassifier_quality_test, average='micro')

    accuracy_rfclassifier_quality=accuracy_score(y_test, y_pred_rfclassifier_quality_test)
    print("Accuracy rf_class: ", accuracy_rfclassifier_quality)
    print('f1_score rf_class(train): ', f1_score_rfclassifier_quality_train)
    print('f1_score rf_class(test): ', f1_score_rfclassifier_quality_test)
    return rfclassifier_quality, accuracy_rfclassifier_quality, f1_score_rfclassifier_quality_train, f1_score_rfclassifier_quality_test, cv_rf_class_quality,y_pred_rfclassifier_quality_test


accuracy_dtclassifier_quality, f1_score_dtclassifier_quality_train, f1_score_dtclassifier_quality_test, cv_dt_class_quality, y_pred_dtclassifier_quality_test = DecisionTreeClassifierModel(X_train, y_train)
print("--------------------------------------")
accuracy_knnclassifier_quality, f1_score_knnclassifier_quality_train, f1_score_knnclassifier_quality_test, cv_knn_class_quality, y_pred_knnclassifier_quality_test = KNNClassifierModel(X_train, y_train)
print("--------------------------------------")
accuracy_nbclassifier_quality, f1_score_nbclassifier_quality_train, f1_score_nbclassifier_quality_test, cv_nb_class_quality, y_pred_nbclassifier_quality_test = NaiveBayesClassifierModel(X_train, y_train)
print("--------------------------------------")
rfclassifier_quality, accuracy_rfclassifier_quality, f1_score_rfclassifier_quality_train, f1_score_rfclassifier_quality_test, cv_rf_class_quality, y_pred_rfclassifier_quality_test = RandomForestClassifierModel(X_train, y_train, None)

Accuracy dt_class: 0.6054256726951919
f1_score dt_class(train):  1.0
f1_score dt_class(test):  0.6054256726951919
--------------------------------------
Accuracy KNN_class:  0.6777679752977503
f1_score KNN_class(train):  0.6965112980996502
f1_score KNN_class(test):  0.6777679752977503
--------------------------------------
Accuracy NB_class:  0.6799735333039258
f1_score NB_class(train):  0.6603006523588919
f1_score NB_class(test):  0.6799735333039258
--------------------------------------
Accuracy rf_class:  0.7051168945743274
f1_score rf_class(train):  1.0
f1_score rf_class(test):  0.7051168945743274


modelsClass_quality = [('Decision Tree Classification', accuracy_dtclassifier_quality, f1_score_dtclassifier_quality_train, f1_score_dtclassifier_quality_test, cv_dt_class_quality.mean() ),
          ('K Nearest Neighbors Classification', accuracy_knnclassifier_quality, f1_score_knnclassifier_quality_train, f1_score_knnclassifier_quality_test, cv_knn_class_quality.mean()),
          ('Naive Bayes', accuracy_nbclassifier_quality, f1_score_nbclassifier_quality_train, f1_score_nbclassifier_quality_test, cv_nb_class_quality.mean()),
          ('Random Forest Classification', accuracy_rfclassifier_quality, f1_score_rfclassifier_quality_train, f1_score_rfclassifier_quality_test, cv_rf_class_quality.mean())
         ]


predictions_class_quality = pd.DataFrame(data = modelsClass_quality, columns=['Model', 'Accuracy', 'F1_Score(training)', 'F1_Score(test)', 'Cross-Validation'])
predictions_class_quality


metrics = ['Accuracy', 'F1_Score(training)', 'F1_Score(test)', 'Cross-Validation']
scores = {metric: [model[i+1] for model in modelsClass_quality] for i, metric in enumerate(metrics)}
models_binaryclassification_list = [model[0] for model in modelsClass_quality]
index = np.arange(len(models_binaryclassification_list))

bar_width = 0.2
plt.figure(figsize=(12, 6))
colors = cm.get_cmap('viridis')(np.linspace(0, 1, len(metrics)))

for i, metric in enumerate(metrics):
    plt.bar(index + i * bar_width, scores[metric], bar_width, label=metric, color=colors[i])
    for j, score in enumerate(scores[metric]):
        plt.text(index[j] + i * bar_width, score + 0.01, round(score, 2), ha='center')

plt.xlabel("Models")
plt.ylabel("Metrics' Scores")
plt.title("Comparison of the different Classification Models' Metrics")
plt.xticks(index + bar_width*1.5, models_binaryclassification_list)
plt.legend()
plt.tight_layout()
plt.show()

The get_cmap function was deprecated in Matplotlib 3.7 and will be removed two minor releases later. Use ``matplotlib.colormaps[name]`` or ``matplotlib.colormaps.get_cmap(obj)`` instead.


fig, axes = plt.subplots(2,2, figsize=(10,8))
skplt.metrics.plot_confusion_matrix(y_test, y_pred_dtclassifier_quality_test, normalize=False, title = 'Confusion Matrix of DecisionTree', ax=axes[0,0])
skplt.metrics.plot_confusion_matrix(y_test, y_pred_knnclassifier_quality_test, normalize=False, title = 'Confusion Matrix of KNN', ax=axes[0,1])
skplt.metrics.plot_confusion_matrix(y_test, y_pred_nbclassifier_quality_test, normalize=False, title = 'Confusion Matrix of Naive Bayes', ax=axes[1,0])
skplt.metrics.plot_confusion_matrix(y_test, y_pred_rfclassifier_quality_test, normalize=False, title = 'Confusion Matrix of RandomForest', ax=axes[1,1])
plt.suptitle("Confusion matrixes of our 4 models")
plt.tight_layout()
plt.show()


#This cell of code took about 40-45min
"""
param_grid = {
    'n_estimators': [10, 50, 100],
    'max_depth': [10, 20, 30],
    'min_samples_split': [2, 5, 10],
    'min_samples_leaf': [1, 2, 4],
    'bootstrap': [True, False]
}

grid_rfclass = RandomForestClassifier()
grid_search = GridSearchCV(estimator = grid_rfclass, param_grid = param_grid, cv = 5, n_jobs = -1)
grid_search.fit(X_train, y_train)

print("Best parameters found : ", grid_search.best_params_)
print("Best estimator found : ", grid_search.best_estimator_)

Outputs :
Best parameters found :  {'bootstrap': False, 'max_depth': 30, 'min_samples_leaf': 4, 'min_samples_split': 10, 'n_estimators': 100}
Best estimator found :  RandomForestClassifier(bootstrap=False, max_depth=30, min_samples_leaf=4,
                       min_samples_split=10)
"""

'\nparam_grid = {\n    \'n_estimators\': [10, 50, 100],\n    \'max_depth\': [10, 20, 30],\n    \'min_samples_split\': [2, 5, 10],\n    \'min_samples_leaf\': [1, 2, 4],\n    \'bootstrap\': [True, False]\n}\n\ngrid_rfclass = RandomForestClassifier()\ngrid_search = GridSearchCV(estimator = grid_rfclass, param_grid = param_grid, cv = 5, n_jobs = -1)\ngrid_search.fit(X_train, y_train)\n\nprint("Best parameters found : ", grid_search.best_params_)\nprint("Best estimator found : ", grid_search.best_estimator_)\n\nOutputs :\nBest parameters found :  {\'bootstrap\': False, \'max_depth\': 30, \'min_samples_leaf\': 4, \'min_samples_split\': 10, \'n_estimators\': 100}\nBest estimator found :  RandomForestClassifier(bootstrap=False, max_depth=30, min_samples_leaf=4,\n                       min_samples_split=10)\n'


#This cell of code took about 30min
"""
param_grid = {
    'n_estimators': [100,200],
    'max_depth': [30,50],
    'min_samples_split': [10,20],
    'min_samples_leaf': [4,5],
}

grid_rfclass = RandomForestClassifier()
grid_search = GridSearchCV(estimator = grid_rfclass, param_grid = param_grid, cv = 5, n_jobs = -1)
grid_search.fit(X_train, y_train)

print("Best parameters found : ", grid_search.best_params_)
print("Best estimator found : ", grid_search.best_estimator_)
print("Score associated : ",grid_search.best_score_)

Outputs :
Best parameters found :  {'max_depth': 30, 'min_samples_leaf': 4, 'min_samples_split': 20, 'n_estimators': 200}
Best estimator found :  RandomForestClassifier(max_depth=30, min_samples_leaf=4, min_samples_split=20,
                       n_estimators=200)
Score associated :  0.688181619066665
"""

'\nparam_grid = {\n    \'n_estimators\': [100,200],\n    \'max_depth\': [30,50],\n    \'min_samples_split\': [10,20],\n    \'min_samples_leaf\': [4,5],\n}\n\ngrid_rfclass = RandomForestClassifier()\ngrid_search = GridSearchCV(estimator = grid_rfclass, param_grid = param_grid, cv = 5, n_jobs = -1)\ngrid_search.fit(X_train, y_train)\n\nprint("Best parameters found : ", grid_search.best_params_)\nprint("Best estimator found : ", grid_search.best_estimator_)\nprint("Score associated : ",grid_search.best_score_)\n\nOutputs :\nBest parameters found :  {\'max_depth\': 30, \'min_samples_leaf\': 4, \'min_samples_split\': 20, \'n_estimators\': 200}\nBest estimator found :  RandomForestClassifier(max_depth=30, min_samples_leaf=4, min_samples_split=20,\n                       n_estimators=200)\nScore associated :  0.688181619066665\n'


#param_grid = {
#    'n_estimators': [200,300],
#    'max_depth': [30],
#    'min_samples_split': [20,40],
#    'min_samples_leaf': [4]
#}

#grid_rfclass = RandomForestClassifier()
#grid_search = GridSearchCV(estimator = grid_rfclass, param_grid = param_grid, cv = 5, n_jobs = -1)
#grid_search.fit(X_train, y_train)

#print("Best parameters found : ", grid_search.best_params_)
#print("Best estimator found : ", grid_search.best_estimator_)
#print("Score associated : ",grid_search.best_score_)

"""
Output :
Best parameters found :  {'max_depth': 30, 'min_samples_leaf': 4, 'min_samples_split': 20, 'n_estimators': 200}
Best estimator found :  RandomForestClassifier(max_depth=30, min_samples_leaf=4, min_samples_split=20,
                       n_estimators=200)
Score associated :  0.6878970582938029
"""

"\nOutput :\nBest parameters found :  {'max_depth': 30, 'min_samples_leaf': 4, 'min_samples_split': 20, 'n_estimators': 200}\nBest estimator found :  RandomForestClassifier(max_depth=30, min_samples_leaf=4, min_samples_split=20,\n                       n_estimators=200)\nScore associated :  0.6878970582938029\n"


param_grid_rf = {
    'n_estimators': 200,
    'max_depth': 30,
    'min_samples_split': 20,
    'min_samples_leaf': 4,
    'random_state': 42
}


rfclassifier, accuracy_rfclassifier, f1_score_rfclassifier_train, f1_score_rfclassifier_test, cv_rf_class, y_pred_rfclassifier_test = RandomForestClassifierModel(X_train, y_train, param_grid_rf)

Accuracy rf_class:  0.7093074547860608
f1_score rf_class(train):  0.8828590337524818
f1_score rf_class(test):  0.7093074547860608


feature_importances = rfclassifier.feature_importances_
feature_names = data_quality.drop(['shares'], axis=1).columns
sorted_indices = np.argsort(feature_importances)[::-1]  # sort the features by their importance

palette = sns.color_palette("viridis", n_colors=len(feature_importances))
plt.figure(figsize=(16, 8))
plt.title("Feature Importances from Random Forest Regression")
plt.barh(range(len(feature_importances)), feature_importances[sorted_indices], color=palette)
plt.yticks(range(len(feature_importances)), feature_names[sorted_indices])
plt.xlabel('Importance')
plt.ylabel('Features')
plt.gca().invert_yaxis()
plt.show()


probas_rf = rfclassifier.predict_proba(X_test)
probas_rf = probas_rf[:,1] #keep the probabilities of the class positive only
fpr, tpr, thresholds = roc_curve(y_test, probas_rf)
roc_auc = auc(fpr, tpr)

plt.figure()
plt.plot(fpr, tpr, color='darkorange', lw=1, label='ROC curve (area = %0.2f)' % roc_auc)
plt.plot([0, 1], [0, 1], color='navy', linestyle='--')
plt.xlim([0.0, 1.0])
plt.ylim([0.0, 1.05])
plt.xlabel('False Positive Rate')
plt.ylabel('True Positive Rate')
plt.title('Receiver Operating Characteristic (ROC) Curve')
plt.legend(loc="lower right")
plt.show()


new_features = feature_names[sorted_indices][:15].append(pd.Index(['shares']))
X_train_topfeatures, X_test_topfeatures, y_train_topfeatures, y_test_topfeatures = SplitAndScaler(data_quality_binary_classes[new_features], 0.3, True)


"""
param_grid_top15 = {
    'n_estimators': [100,200],
    'max_depth': [30,50],
    'min_samples_split': [5,10],
    'min_samples_leaf': [4,5]
}

grid_rfclass_quality_top15 = RandomForestClassifier()
grid_search_top15 = GridSearchCV(estimator = grid_rfclass_quality_top15, param_grid = param_grid_top15, cv = 5, n_jobs = -1)
grid_search_top15.fit(X_train_topfeatures, y_train_topfeatures)

print("Best parameters found : ", grid_search_top15.best_params_)
print("Score associated : ",grid_search_top15.best_score_)

Output :
Best parameters found :  {'max_depth': 50, 'min_samples_leaf': 5, 'min_samples_split': 5, 'n_estimators': 200}
Score associated :  0.6893212367678287
"""

'\nparam_grid_top15 = {\n    \'n_estimators\': [100,200],\n    \'max_depth\': [30,50],\n    \'min_samples_split\': [5,10],\n    \'min_samples_leaf\': [4,5]\n}\n\ngrid_rfclass_quality_top15 = RandomForestClassifier()\ngrid_search_top15 = GridSearchCV(estimator = grid_rfclass_quality_top15, param_grid = param_grid_top15, cv = 5, n_jobs = -1)\ngrid_search_top15.fit(X_train_topfeatures, y_train_topfeatures)\n\nprint("Best parameters found : ", grid_search_top15.best_params_)\nprint("Score associated : ",grid_search_top15.best_score_)\n\nOutput :\nBest parameters found :  {\'max_depth\': 50, \'min_samples_leaf\': 5, \'min_samples_split\': 5, \'n_estimators\': 200}\nScore associated :  0.6893212367678287\n'


rfclass_quality_top15 = RandomForestClassifier(n_estimators=200,max_depth=50,min_samples_split=5,min_samples_leaf=5,random_state=42)
rfclass_quality_top15.fit(X_train_topfeatures, y_train_topfeatures)

cv_rfclass_quality_top15 = cross_val_score(estimator = rfclass_quality_top15, X = X_train_topfeatures, y = y_train_topfeatures, cv = 6)

y_pred_rfclass_quality_top15_train = rfclass_quality_top15.predict(X_train_topfeatures)
f1_score_rfclass_quality_top15_train = f1_score(y_train_topfeatures, y_pred_rfclass_quality_top15_train, average='micro')

y_pred_rfclass_quality_top15_test = rfclass_quality_top15.predict(X_test_topfeatures)
f1_score_rfclass_quality_top15_test = f1_score(y_test_topfeatures, y_pred_rfclass_quality_top15_test, average='micro')

accuracy_rfclass_quality_top15=accuracy_score(y_test_topfeatures, y_pred_rfclass_quality_top15_test)
print("Accuracy rf_class: ", accuracy_rfclass_quality_top15)
print('f1_score rf_class(train): ', f1_score_rfclass_quality_top15_train)
print('f1_score rf_class(test): ', f1_score_rfclass_quality_top15_test)

Accuracy rf_class:  0.7011468901632113
f1_score rf_class(train):  0.9191642242601872
f1_score rf_class(test):  0.7011468901632113


data_quality_multi_classes = data_quality.copy()
data_quality_multi_classes_plot = data_quality.copy()
multi_labels, bins = pd.qcut(data_quality_multi_classes['shares'], q=5, labels=False, retbins=True)
data_quality_multi_classes['shares'] = multi_labels
data_quality_multi_classes_plot['popularity'] = multi_labels


X_train, X_test, y_train, y_test = SplitAndScaler(data_quality_multi_classes, 0.3, True)


accuracy_dtclassifier_quality, f1_score_dtclassifier_quality_train, f1_score_dtclassifier_quality_test, cv_dt_class_quality, y_pred_dtclassifier_quality_test = DecisionTreeClassifierModel(X_train, y_train)
print("--------------------------------------")
accuracy_knnclassifier_quality, f1_score_knnclassifier_quality_train, f1_score_knnclassifier_quality_test, cv_knn_class_quality, y_pred_knnclassifier_quality_test = KNNClassifierModel(X_train, y_train)
print("--------------------------------------")
accuracy_nbclassifier_quality, f1_score_nbclassifier_quality_train, f1_score_nbclassifier_quality_test, cv_nb_class_quality, y_pred_nbclassifier_quality_test = NaiveBayesClassifierModel(X_train, y_train)
print("--------------------------------------")
rfclassifier_quality, accuracy_rfclassifier_quality, f1_score_rfclassifier_quality_train, f1_score_rfclassifier_quality_test, cv_rf_class_quality, y_pred_rfclassifier_quality_test = RandomForestClassifierModel(X_train, y_train, None)

Accuracy dt_class: 0.2516541685046317
f1_score dt_class(train):  1.0
f1_score dt_class(test):  0.2516541685046317
--------------------------------------
Accuracy KNN_class:  0.27591530657256286
f1_score KNN_class(train):  0.3665500614540985
f1_score KNN_class(test):  0.27591530657256286
--------------------------------------
Accuracy NB_class:  0.30149977944419937
f1_score NB_class(train):  0.29857237401909803
f1_score NB_class(test):  0.30149977944419937
--------------------------------------
Accuracy rf_class:  0.3266431407146008
f1_score rf_class(train):  1.0
f1_score rf_class(test):  0.3266431407146008


modelsMultiClass_quality = [('Decision Tree Classification', accuracy_dtclassifier_quality, f1_score_dtclassifier_quality_train, f1_score_dtclassifier_quality_test, cv_dt_class_quality.mean() ),
          ('K Nearest Neighbors Classification', accuracy_knnclassifier_quality, f1_score_knnclassifier_quality_train, f1_score_knnclassifier_quality_test, cv_knn_class_quality.mean()),
          ('Naive Bayes', accuracy_nbclassifier_quality, f1_score_nbclassifier_quality_train, f1_score_nbclassifier_quality_test, cv_nb_class_quality.mean()),
          ('Random Forest Classification', accuracy_rfclassifier_quality, f1_score_rfclassifier_quality_train, f1_score_rfclassifier_quality_test, cv_rf_class_quality.mean())
         ]


predictions_multiclass_quality = pd.DataFrame(data = modelsMultiClass_quality, columns=['Model', 'Accuracy', 'F1_Score(training)', 'F1_Score(test)', 'Cross-Validation'])
predictions_multiclass_quality


metrics = ['Accuracy', 'F1_Score(training)', 'F1_Score(test)', 'Cross-Validation']
scores = {metric: [model[i+1] for model in modelsMultiClass_quality] for i, metric in enumerate(metrics)}
models_multiclassification_list = [model[0] for model in modelsMultiClass_quality]
index = np.arange(len(models_multiclassification_list))

bar_width = 0.2
plt.figure(figsize=(12, 6))

colors = cm.get_cmap('viridis')(np.linspace(0, 1, len(metrics)))

for i, metric in enumerate(metrics):
    plt.bar(index + i * bar_width, scores[metric], bar_width, label=metric, color=colors[i])
    for j, score in enumerate(scores[metric]):
        plt.text(index[j] + i * bar_width, score + 0.01, round(score, 2), ha='center')

plt.xlabel("Models")
plt.ylabel("Metrics' Scores")
plt.title("Comparison of the different Classification Models' Metrics")
plt.xticks(index + bar_width*1.5, models_multiclassification_list)
plt.legend()
plt.tight_layout()
plt.show()

The get_cmap function was deprecated in Matplotlib 3.7 and will be removed two minor releases later. Use ``matplotlib.colormaps[name]`` or ``matplotlib.colormaps.get_cmap(obj)`` instead.


fig, axes = plt.subplots(2,2, figsize=(10,8))
skplt.metrics.plot_confusion_matrix(y_test, y_pred_dtclassifier_quality_test, normalize=False, title = 'Confusion Matrix of DecisionTree', ax=axes[0,0])
skplt.metrics.plot_confusion_matrix(y_test, y_pred_knnclassifier_quality_test, normalize=False, title = 'Confusion Matrix of KNN', ax=axes[0,1])
skplt.metrics.plot_confusion_matrix(y_test, y_pred_nbclassifier_quality_test, normalize=False, title = 'Confusion Matrix of Naive Bayes', ax=axes[1,0])
skplt.metrics.plot_confusion_matrix(y_test, y_pred_rfclassifier_quality_test, normalize=False, title = 'Confusion Matrix of RandomForest', ax=axes[1,1])
plt.suptitle("Confusion matrixes of our 4 models")
plt.tight_layout()
plt.show()


data_quality_binary_classes = data_quality.copy()
print(data_quality_binary_classes['shares'].mean())
popular=data_quality_binary_classes['shares']>=data_quality_binary_classes['shares'].mean()
unpopular=data_quality_binary_classes['shares']<data_quality_binary_classes['shares'].mean()
data_quality_binary_classes.loc[popular,'shares']=1
data_quality_binary_classes.loc[unpopular,'shares']=0
data_quality_binary_classes['shares'].value_counts()

1540.7552114353782

0    9996
1    5115
Name: shares, dtype: int64


X_train, X_test, y_train, y_test = SplitAndScaler(data_quality_binary_classes, 0.2, True)
validation_size=int(X_test.shape[0])
X_val=X_train[:validation_size]
X_train=X_train[validation_size:]

y_val=y_train[:validation_size]
y_train=y_train[validation_size:]
print("X_val",X_val.shape)
print("X_test",X_test.shape)
print("y_val",y_val.shape)
print("y_test",y_test.shape)
print("X_train",X_train.shape)
print("y_train",y_train.shape)
print("train+set+val=",X_val.shape[0]+X_test.shape[0]+X_train.shape[0])
print("'data_quality_binary_classes' rows",data_quality_binary_classes.shape[0])

X_val (3023, 24)
X_test (3023, 24)
y_val (3023,)
y_test (3023,)
X_train (9065, 24)
y_train (9065,)
train+set+val= 15111
'data_quality_binary_classes' rows 15111


y_train = to_categorical(y_train)
y_test = to_categorical(y_test)
y_val = to_categorical(y_val)


ann_binary = Sequential()

ann_binary.add(Dense(input_shape = (24,), units = 256, activation='relu'))
ann_binary.add(Dropout(0.3)) #we add Dropout because the model was overfitting
ann_binary.add(BatchNormalization())
ann_binary.add(Dense(units=128, activation='relu'))
ann_binary.add(Dropout(0.3))
ann_binary.add(Dense(units=64, activation='relu'))
ann_binary.add(Dropout(0.3))
ann_binary.add(Dense(units=32, activation='relu'))
ann_binary.add(Dense(units=2, activation='sigmoid'))

WARNING:tensorflow:From C:\Users\sebmo\anaconda3\lib\site-packages\keras\src\backend.py:873: The name tf.get_default_graph is deprecated. Please use tf.compat.v1.get_default_graph instead.


ann_binary.compile(optimizer = 'adam', loss = 'binary_crossentropy', metrics = ['accuracy'])

early_stop = EarlyStopping(monitor='val_loss', patience=2)
history = ann_binary.fit(X_train, y_train, validation_data=(X_val, y_val), epochs=20, callbacks=[early_stop])

WARNING:tensorflow:From C:\Users\sebmo\anaconda3\lib\site-packages\keras\src\optimizers\__init__.py:309: The name tf.train.Optimizer is deprecated. Please use tf.compat.v1.train.Optimizer instead.

Epoch 1/20
WARNING:tensorflow:From C:\Users\sebmo\anaconda3\lib\site-packages\keras\src\utils\tf_utils.py:492: The name tf.ragged.RaggedTensorValue is deprecated. Please use tf.compat.v1.ragged.RaggedTensorValue instead.

WARNING:tensorflow:From C:\Users\sebmo\anaconda3\lib\site-packages\keras\src\engine\base_layer_utils.py:384: The name tf.executing_eagerly_outside_functions is deprecated. Please use tf.compat.v1.executing_eagerly_outside_functions instead.

284/284 [==============================] - 3s 5ms/step - loss: 0.6418 - accuracy: 0.6491 - val_loss: 0.6175 - val_accuracy: 0.6619
Epoch 2/20
284/284 [==============================] - 1s 4ms/step - loss: 0.6121 - accuracy: 0.6726 - val_loss: 0.6063 - val_accuracy: 0.6705
Epoch 3/20
284/284 [==============================] - 1s 4ms/step - loss: 0.5990 - accuracy: 0.6834 - val_loss: 0.6024 - val_accuracy: 0.6755
Epoch 4/20
284/284 [==============================] - 1s 4ms/step - loss: 0.5978 - accuracy: 0.6845 - val_loss: 0.5989 - val_accuracy: 0.6775
Epoch 5/20
284/284 [==============================] - 1s 3ms/step - loss: 0.5918 - accuracy: 0.6887 - val_loss: 0.5991 - val_accuracy: 0.6742
Epoch 6/20
284/284 [==============================] - 1s 3ms/step - loss: 0.5883 - accuracy: 0.6927 - val_loss: 0.5968 - val_accuracy: 0.6768
Epoch 7/20
284/284 [==============================] - 1s 5ms/step - loss: 0.5832 - accuracy: 0.6928 - val_loss: 0.6025 - val_accuracy: 0.6742
Epoch 8/20
284/284 [==============================] - 1s 4ms/step - loss: 0.5839 - accuracy: 0.6945 - val_loss: 0.5957 - val_accuracy: 0.6795
Epoch 9/20
284/284 [==============================] - 1s 5ms/step - loss: 0.5761 - accuracy: 0.6994 - val_loss: 0.5988 - val_accuracy: 0.6718
Epoch 10/20
284/284 [==============================] - 1s 4ms/step - loss: 0.5728 - accuracy: 0.7054 - val_loss: 0.5994 - val_accuracy: 0.6758


ann_binary.summary()

Model: "sequential"
_________________________________________________________________
 Layer (type)                Output Shape              Param #   
=================================================================
 dense (Dense)               (None, 256)               6400      
                                                                 
 dropout (Dropout)           (None, 256)               0         
                                                                 
 batch_normalization (Batch  (None, 256)               1024      
 Normalization)                                                  
                                                                 
 dense_1 (Dense)             (None, 128)               32896     
                                                                 
 dropout_1 (Dropout)         (None, 128)               0         
                                                                 
 dense_2 (Dense)             (None, 64)                8256      
                                                                 
 dropout_2 (Dropout)         (None, 64)                0         
                                                                 
 dense_3 (Dense)             (None, 32)                2080      
                                                                 
 dense_4 (Dense)             (None, 2)                 66        
                                                                 
=================================================================
Total params: 50722 (198.13 KB)
Trainable params: 50210 (196.13 KB)
Non-trainable params: 512 (2.00 KB)
_________________________________________________________________


accuracy_train = ann_binary.evaluate(X_train, y_train, verbose=0)
accuracy_test = ann_binary.evaluate(X_test, y_test, verbose=0)
print(f"Binary Crossentropy Train : {round(accuracy_train[0],3)} || Accuracy Train : {round(accuracy_train[1],3)}")
print(f"Binary Crossentropy Test  : {round(accuracy_test[0],3)} || Accuracy Test : {round(accuracy_test[1],3)}")

Binary Crossentropy Train : 0.546 || Accuracy Train : 0.719
Binary Crossentropy Test  : 0.604 || Accuracy Test : 0.69


history.history.keys()

dict_keys(['loss', 'accuracy', 'val_loss', 'val_accuracy'])


plt.plot(history.history['loss'])
plt.plot(history.history['val_loss'])
plt.title('Training Loss Vs Validation Loss')
plt.ylabel('Loss')
plt.xlabel('Epoch')
plt.legend(['train_loss', 'val_loss'], loc='upper left')
plt.show()


plt.plot(history.history['accuracy'])
plt.plot(history.history['val_accuracy'])
plt.title('Training Accuracy Vs Validation Accuracy')
plt.ylabel('Accuracy')
plt.xlabel('Epoch')
plt.legend(['train_accuracy', 'val_accuracy'], loc='upper left')
plt.show()


data_quality_multi_classes = data_quality.copy()
data_quality_multi_classes_plot = data_quality.copy()
multi_labels, bins = pd.qcut(data_quality_multi_classes['shares'], q=5, labels=False, retbins=True)
data_quality_multi_classes['shares'] = multi_labels
data_quality_multi_classes_plot['popularity'] = multi_labels


X_train, X_test, y_train, y_test = SplitAndScaler(data_quality_multi_classes, 0.3, True)
validation_size=int(X_test.shape[0])
X_val=X_train[:validation_size]
X_train=X_train[validation_size:]

y_val=y_train[:validation_size]
y_train=y_train[validation_size:]


y_train = to_categorical(y_train)
y_test = to_categorical(y_test)
y_val = to_categorical(y_val)


ann_multi = Sequential()

ann_multi.add(Dense(input_shape = (24,), units = 256, activation='relu'))
ann_binary.add(Dropout(0.3))
ann_binary.add(BatchNormalization())
ann_binary.add(Dropout(0.3))
ann_multi.add(Dense(units=64, activation='relu'))
ann_binary.add(Dropout(0.3))
ann_multi.add(Dense(units = 5, activation='softmax'))


ann_multi.compile(loss = 'categorical_crossentropy', optimizer='adam', metrics = ['accuracy'])

ann_multi.summary()
early_stop = EarlyStopping(monitor='val_loss', patience=2)
history = ann_multi.fit(X_train, y_train, validation_data=(X_val, y_val), epochs=20, callbacks=[early_stop])

Model: "sequential_1"
_________________________________________________________________
 Layer (type)                Output Shape              Param #   
=================================================================
 dense_5 (Dense)             (None, 256)               6400      
                                                                 
 dense_6 (Dense)             (None, 64)                16448     
                                                                 
 dense_7 (Dense)             (None, 5)                 325       
                                                                 
=================================================================
Total params: 23173 (90.52 KB)
Trainable params: 23173 (90.52 KB)
Non-trainable params: 0 (0.00 Byte)
_________________________________________________________________
Epoch 1/20
189/189 [==============================] - 1s 3ms/step - loss: 1.5643 - accuracy: 0.2815 - val_loss: 1.5416 - val_accuracy: 0.3101
Epoch 2/20
189/189 [==============================] - 1s 3ms/step - loss: 1.5099 - accuracy: 0.3326 - val_loss: 1.5400 - val_accuracy: 0.3163
Epoch 3/20
189/189 [==============================] - 1s 3ms/step - loss: 1.4791 - accuracy: 0.3500 - val_loss: 1.5420 - val_accuracy: 0.3008
Epoch 4/20
189/189 [==============================] - 1s 3ms/step - loss: 1.4574 - accuracy: 0.3657 - val_loss: 1.5529 - val_accuracy: 0.3022


multi_accuracy_train = ann_multi.evaluate(X_train, y_train, verbose=0)
multi_accuracy_test = ann_multi.evaluate(X_test, y_test, verbose=0)
print(f"Categorical Crossentropy Train : {round(multi_accuracy_train[0],3)} || Accuracy Train : {round(multi_accuracy_train[1],3)}")
print(f"Categorical Crossentropy Test  : {round(multi_accuracy_test[0],3)} || Accuracy Test : {round(multi_accuracy_test[1],3)}")

Categorical Crossentropy Train : 1.41 || Accuracy Train : 0.406
Categorical Crossentropy Test  : 1.548 || Accuracy Test : 0.299


plt.plot(history.history['loss'])
plt.plot(history.history['val_loss'])
plt.title('Training Loss Vs Validation Loss')
plt.ylabel('Loss')
plt.xlabel('Epoch')
plt.legend(['train_loss', 'val_loss'], loc='upper left')
plt.show()


plt.plot(history.history['accuracy'])
plt.plot(history.history['val_accuracy'])
plt.title('Training Accuracy Vs Validation Accuracy')
plt.ylabel('Accuracy')
plt.xlabel('Epoch')
plt.legend(['train_accuracy', 'val_accuracy'], loc='upper left')
plt.show()


avg_shares_by_title_length = df.groupby('n_tokens_title')['shares'].mean().reset_index()

fig1 = px.bar(avg_shares_by_title_length, x='n_tokens_title', y='shares', title='Number of Words in Title in relation with Shares',
              labels={'n_tokens_title': 'Number of Words in Title', 'shares': 'Average Number of Shares'})

fig1.update_layout(autosize=False, width=600, height=400)
fig1.show()


bins = list(range(0, 3500, 200))
df['content_length_bins'] = pd.cut(df['n_tokens_content'], bins=bins).astype(str)
order = [f'({bins[i]}, {bins[i+1]}]' for i in range(len(bins)-1)]

avg_shares_by_content_length = df.groupby('content_length_bins', as_index=False)['shares'].mean()
avg_shares_by_content_length['content_length_bins'] = pd.Categorical(avg_shares_by_content_length['content_length_bins'], categories=order, ordered=True)
avg_shares_by_content_length = avg_shares_by_content_length.sort_values(by='content_length_bins')

max_index = avg_shares_by_content_length['shares'].idxmax()

fig = px.line(avg_shares_by_content_length, x='content_length_bins', y='shares', title='Average Shares by Content Length',
              labels={'content_length_bins': 'Content Length Range', 'shares': 'Average Number of Shares'})

fig.update_traces(line=dict(color='blue', shape='linear'), marker=dict(color='green'), selector=dict(type='scatter'))

fig.add_scatter(x=[avg_shares_by_content_length.loc[max_index, 'content_length_bins']],
                y=[avg_shares_by_content_length.loc[max_index, 'shares']],
                mode='markers',
                marker=dict(color='green', size=10),
                showlegend=False)

fig.update_layout(autosize=False, width=800, height=400)
fig.show()


df['n_unique_tokens_bins'] = pd.cut(df['n_unique_tokens'], bins=np.linspace(0, 1, 11), labels=False) * 0.1
avg_shares_by_bins = df.groupby('n_unique_tokens_bins')['shares'].mean().reset_index()

plt.figure(figsize=(12, 8))
plt.stem(avg_shares_by_bins['n_unique_tokens_bins'], avg_shares_by_bins['shares'], basefmt='k-', use_line_collection=True)
plt.scatter(avg_shares_by_bins['n_unique_tokens_bins'], avg_shares_by_bins['shares'], color='red', s=50)
plt.title('Average Shares per Interval of Unique Words Ratio', fontsize=16)
plt.xlabel('Unique Words Ratio Interval', fontsize=14)
plt.ylabel('Average Number of Shares', fontsize=14)

plt.xticks(np.linspace(0, 0.9, 10), [f'{i:.1f}-{i+0.1:.1f}' for i in np.linspace(0, 0.9, 10)], rotation=45)

for i, txt in enumerate(avg_shares_by_bins['shares']):
    plt.text(avg_shares_by_bins['n_unique_tokens_bins'][i], txt + 20, f'{int(txt)}', ha='center', va='bottom', fontsize=10, color='blue')

plt.show()

C:\Users\sebmo\AppData\Local\Temp\ipykernel_676\984228224.py:5: MatplotlibDeprecationWarning:

The 'use_line_collection' parameter of stem() was deprecated in Matplotlib 3.6 and will be removed two minor releases later. If any parameter follows 'use_line_collection', they should be passed as keyword, not positionally.


df['average_token_length'].describe()

count    15111.000000
mean         4.635346
std          0.217294
min          3.653846
25%          4.492308
50%          4.657175
75%          4.805310
max          5.000000
Name: average_token_length, dtype: float64


df['average_token_length'] = df['average_token_length'].apply(lambda x: min(x, 8))


fig = px.scatter(df, x='average_token_length', y='shares')
fig.update_layout(
    width=1000,
    height=450
)
fig.show()


df['color'] = ['red' if neg > pos else 'blue' for neg, pos in zip(df['global_rate_negative_words'], df['global_rate_positive_words'])]

fig = px.scatter_3d(df, x='global_rate_positive_words', y='global_rate_negative_words', z='shares', color='color', size_max=30, opacity=0.7)

fig.update_layout(title='3D Plot of Positive and Negative Word Rates vs. Shares',
                  scene=dict(xaxis_title='Rate of Positive Words', yaxis_title='Rate of Negative Words', zaxis_title='Shares'),
                  showlegend=False, height=600)

fig.show()


import seaborn as sns
import matplotlib.pyplot as plt
from plotly.subplots import make_subplots
import plotly.graph_objects as go


fig = make_subplots(rows=1, cols=2, subplot_titles=['Positive Word Rate vs. Shares', 'Negative Word Rate vs. Shares'])
scatter1 = go.Scatter(x=df['global_rate_positive_words'], y=df['shares'], mode='markers', marker=dict(color=df['shares'], size=10),
                      showlegend=False)
fig.add_trace(scatter1, row=1, col=1)
scatter2 = go.Scatter(x=df['global_rate_negative_words'], y=df['shares'], mode='markers', marker=dict(color=df['shares'], size=10),
                      showlegend=False)
fig.add_trace(scatter2, row=1, col=2)
fig.update_layout(title_text='Scatter Plots of Positive and Negative Word Rates vs. Shares',
                  xaxis=dict(title='Rate of Words'),
                  yaxis=dict(title='Number of Shares'),
                  showlegend=False,
                 height=600)
fig.show()

filtered_data_positive = df[df['global_rate_positive_words'] > df['global_rate_negative_words']]
filtered_data_negative = df[df['global_rate_negative_words'] > df['global_rate_positive_words']]

plt.figure(figsize=(12, 8))
sns.violinplot(x=filtered_data_positive['shares'], color='skyblue')
ax = sns.violinplot(x=filtered_data_negative['shares'], color='orange')
plt.title('Distribution of Shares for rate_positive_words and rate_negative_words > 0.5')
plt.xlabel('Number of Shares')
plt.ylabel('Density')
legend_labels = ['Blue: rate_positive_words > 0.5 (Positive)', 'Orange: rate_negative_words > 0.5 (Negative)']
ax.legend(labels=legend_labels)

plt.show()


global_pos_rate_neg = df[(df['global_sentiment_polarity'] > 0) & (df['global_rate_negative_words'] > df['global_rate_positive_words'])]

plt.figure(figsize=(10, 6))
plt.scatter(global_pos_rate_neg['global_sentiment_polarity'], global_pos_rate_neg['global_rate_negative_words'],
            color='red', marker='o', label='Filtered Rows')

plt.title('Scatter Plot of Global Sentiment Polarity vs. Rate of Negative Words')
plt.xlabel('Global Sentiment Polarity')
plt.ylabel('Rate of Negative Words')
plt.show()


filtered_data_positive = df[df['global_rate_positive_words'] > df['global_rate_negative_words']]

point_size = filtered_data_positive['shares'] / filtered_data_positive['shares'].max() * 200

plt.figure(figsize=(12, 8))
hb = plt.hexbin(filtered_data_positive['avg_positive_polarity'], filtered_data_positive['avg_negative_polarity'],
                C=filtered_data_positive['shares'], gridsize=50, cmap='cool', mincnt=1, edgecolors='gray', linewidths=0.2)

plt.title('Relation between Avg Positive Polarity, Avg Negative Polarity et Shares (rate_positive_words > 0.5)')
plt.xlabel('Average Positive Polarity')
plt.ylabel('Average Negative Polarity')
plt.colorbar(hb, label='Number of Shares')
plt.show()


bins_positive = [0, 0.33, 0.67, 1]
labels_positive = ['slightly positive', 'positive', 'very positive']
bins_negative = [-1, -0.67, -0.33, 0]
labels_negative = ['very negative', 'negative', 'slightly negative']

df['positive_sentiment_category'] = pd.cut(df['global_sentiment_polarity'], bins=bins_positive, labels=labels_positive, right=False)
df['negative_sentiment_category'] = pd.cut(df['global_sentiment_polarity'], bins=bins_negative, labels=labels_negative, right=False)

positive_data = df.groupby('positive_sentiment_category')['shares'].mean().reset_index()
negative_data = df.groupby('negative_sentiment_category')['shares'].mean().reset_index()

merged_data = pd.merge(positive_data, negative_data, how='outer', left_on='positive_sentiment_category', right_on='negative_sentiment_category', suffixes=('_positive', '_negative'))
merged_data['positive_sentiment_category'] = merged_data['positive_sentiment_category'].astype(str)

melted_df = pd.melt(merged_data, id_vars=['positive_sentiment_category', 'negative_sentiment_category'],
                    value_vars=['shares_positive', 'shares_negative'],
                    var_name='sentiment_type', value_name='shares')
melted_df['positive_sentiment_category'] = melted_df['positive_sentiment_category'].astype(str)
melted_df['negative_sentiment_category'] = melted_df['negative_sentiment_category'].astype(str)
melted_df['sentiment_category'] = (melted_df['positive_sentiment_category'].replace('nan', '') + ' ' + melted_df['negative_sentiment_category'].replace('nan', ''))
melted_df = melted_df.dropna(subset=['shares'])
melted_df = melted_df[['sentiment_category', 'shares']]
melted_df = melted_df.sort_values(by='shares', ascending=False)

fig, ax = plt.subplots(figsize=(12, 8), dpi=80)
ax.hlines(y=melted_df['sentiment_category'], xmin=0, xmax=melted_df['shares'], color='gray', alpha=0.7, linewidth=1, linestyles='dashdot')
ax.scatter(y=melted_df['sentiment_category'], x=melted_df['shares'], s=75, color='green', alpha=0.7)
ax.set_title('Average Number of Shares by Sentiment Category', fontdict={'size': 16})
ax.set_xlabel('Average Number of Shares')
ax.set_yticks(melted_df['sentiment_category'])
ax.set_yticklabels(melted_df['sentiment_category'].str.title(), fontdict={'horizontalalignment': 'right'})
plt.show()


import plotly.express as px

fig = px.scatter(df, x='global_subjectivity', y='shares', color='shares',
                 color_continuous_scale='cividis', title='Content Subjectivity Score In Relation To Shares',
                 size='shares', size_max=10)

fig.update_layout(xaxis_title='global subjectivity', height=600)

fig.show()


fig = px.scatter(df, x='title_subjectivity', y='shares', color='shares',
                 color_continuous_scale='solar', title='Title Subjectivity Score In Relation To Shares')
fig.update_layout(xaxis_title='title subjectivity', height=600) 
fig.show()


df["weekday_names"] = df["weekday"].astype(int).map({
    1: 'Monday',
    2: 'Tuesday',
    3: 'Wednesday',
    4: 'Thursday',
    5: 'Friday',
    6: 'Saturday',
    7: 'Sunday'
})


grouped_data = df.groupby('weekday_names').agg({'weekday_names': 'count', 'shares': 'mean'}).rename(columns={'weekday_names': 'count'}).reset_index()

fig = px.treemap(grouped_data, path=['weekday_names'], values='count', color='shares', hover_data=['shares'],
                 color_continuous_scale='cividis', title='Count and Average Shares by Day')
fig.show()


average_shares_per_theme = df.groupby('theme')['shares'].mean().reset_index()

fig = go.Figure(go.Pie(
    labels=average_shares_per_theme['theme'],
    values=average_shares_per_theme['shares'],
    pull=[0, 0, 0, 0, 0.2,0,0],textinfo='label+percent'),layout = go.Layout(height=550, width=700))
fig.update_layout(
    title='Average Number of Shares per Theme')
fig.show()


num_themes = df['theme'].nunique()
colorscale = 'Viridis'
theme_group = df.groupby('theme')['shares'].mean().sort_values(ascending=True)
fig = go.Figure()
fig.add_trace(go.Bar(
    y=theme_group.index,
    x=theme_group.values,
    orientation='h',
    marker=dict(color=theme_group.values, colorscale=colorscale)))
fig.update_layout(
    title='Average Shares by Articles Themes',
    xaxis=dict(title='Average Popularity'),
    yaxis=dict(title='Themes'))
fig.show()


df_grouped = df.groupby('num_imgs')['shares'].agg(['mean', 'count']).reset_index()
df_grouped['percentage'] = (df_grouped['count'] / df_grouped['count'].sum()) * 100
df_grouped['percentage']= df_grouped['percentage'].round(2)
df_grouped['percentage']= df_grouped['percentage'].astype(str) + "%"

fig = px.bar(df_grouped, x='mean', y='num_imgs', orientation='h',text='percentage', title='Average Shares and Percentage by Number of Images',labels={'mean': 'Average Shares', 'num_imgs': 'Number of Images'},color='mean',color_continuous_scale='Oranges')
fig.update_layout(yaxis=dict(tickmode='linear', dtick=1))
fig.show()


df_grouped = df.groupby('num_videos')['shares'].agg(['mean', 'count']).reset_index()
df_grouped['percentage'] = (df_grouped['count'] / df_grouped['count'].sum()) * 100
df_grouped['percentage']= df_grouped['percentage'].round(2)
df_grouped['percentage']= df_grouped['percentage'].astype(str) + "%"

fig = px.bar(df_grouped, x='num_videos', y='mean', orientation='v',text='percentage', title='Average Shares and Percentage by Number of Videos',labels={'mean': 'Average Shares', 'num_imgs': 'Number of Videos'},color='mean',color_continuous_scale='Blues')
fig.show()


fig = px.scatter(df.groupby('num_keywords')['shares'].mean().reset_index(), x='num_keywords', y='shares', size='shares', color='shares', title='Average Shares in Relation With Number of Keywords',
                 labels={'num_keywords': 'Number of Keywords', 'shares': 'Average Shares'},
                 hover_name='shares',
                 color_continuous_scale='Oryel')
fig.show()


grouped_data = df.groupby('num_hrefs')['shares'].mean().reset_index()

# Create a bar plot with larger bars
fig = px.bar(grouped_data, x='shares', y='num_hrefs', orientation='h', color='shares',
             labels={'num_hrefs': 'Number of Hrefs', 'shares': 'Average Shares'},
             title='Average Shares by Number of Hrefs',
             color_discrete_sequence=['orange'],
             width=800, height=600)

# Show the plot
fig.show()


grouped_data = df.groupby('num_self_hrefs')['shares'].mean().reset_index()

fig = px.scatter(grouped_data, x='num_self_hrefs', y='shares', size='num_self_hrefs', color='shares',
                 size_max=30, title='Average Shares in relation with the Number of Internal Links',
                 labels={'num_self_hrefs': 'Number of Internal Links', 'shares': 'Average Shares'})

fig.update_traces(marker=dict(line=dict(color='white', width=0.5)))
fig.show()


bins = list(range(0, 3500, 200))
df['range_values'] = pd.cut(df['self_reference_avg_sharess'], bins=bins).astype(str)
order = [f'({bins[i]}, {bins[i+1]}]' for i in range(len(bins)-1)]

avg_shares_by_self_reference = df.groupby('range_values', as_index=False)['shares'].mean()
avg_shares_by_self_reference['range_values'] = pd.Categorical(avg_shares_by_self_reference['range_values'], categories=order, ordered=True)
avg_shares_by_self_reference = avg_shares_by_self_reference.sort_values(by='range_values')

max_index = avg_shares_by_self_reference['shares'].idxmax()

fig = px.line(avg_shares_by_self_reference, x='range_values', y='shares', title='Average Shares by Content Length',
              labels={'range_values': 'Content Length Range', 'shares': 'Average Number of Shares'})

fig.update_traces(line=dict(color='blue', shape='linear'), marker=dict(color='green'), selector=dict(type='scatter'))

fig.add_scatter(x=[avg_shares_by_self_reference.loc[max_index, 'range_values']],
                y=[avg_shares_by_self_reference.loc[max_index, 'shares']],
                mode='markers',
                marker=dict(color='green', size=10),
                showlegend=False)

fig.update_layout(autosize=False, width=800, height=400)
fig.show()

	kw_min_min	kw_max_min	kw_avg_min	kw_min_max	kw_max_max	kw_avg_max	kw_min_avg	kw_max_avg	kw_avg_avg	num_keywords
count	35103.000000	35103.000000	35103.000000	35103.000000	35103.000000	35103.000000	35103.000000	35103.000000	35103.000000	35103.000000
mean	26.227758	1121.656082	306.713224	13355.270718	751072.791499	256557.562416	1087.880379	5460.246489	3055.375204	7.199385
std	69.777677	3531.709537	562.447968	57972.540327	216275.867467	133850.741969	1112.085528	5606.053659	1229.135369	1.916833
min	-1.000000	0.000000	-1.000000	0.000000	0.000000	0.000000	-1.000000	0.000000	0.000000	1.000000
25%	-1.000000	444.000000	140.732222	0.000000	843300.000000	171815.476191	0.000000	3530.699709	2351.233028	6.000000
50%	-1.000000	654.000000	234.285714	1400.000000	843300.000000	241825.000000	1002.000000	4230.133333	2815.495983	7.000000
75%	4.000000	1000.000000	355.285714	7400.000000	843300.000000	327461.904762	1991.985437	5851.933367	3489.782146	9.000000
max	377.000000	298400.000000	39979.000000	843300.000000	843300.000000	843300.000000	3610.124972	298400.000000	37607.521654	10.000000

	num_hrefs	num_self_hrefs	self_reference_min_shares	self_reference_max_shares	self_reference_avg_sharess
count	35103.000000	35103.000000	35103.000000	35103.000000	35103.000000
mean	10.584423	3.297126	3609.750298	9563.299056	5861.762346
std	10.771551	3.847926	19058.826160	39430.352206	23389.533674
min	0.000000	0.000000	0.000000	0.000000	0.000000
25%	4.000000	1.000000	622.000000	1000.000000	952.250000
50%	7.000000	3.000000	1200.000000	2700.000000	2100.000000
75%	13.000000	4.000000	2400.000000	7400.000000	4753.954545
max	187.000000	116.000000	843300.000000	843300.000000	843300.000000

	global_sentiment_polarity	avg_positive_polarity	min_positive_polarity	max_positive_polarity	avg_negative_polarity	min_negative_polarity	max_negative_polarity	title_sentiment_polarity	abs_title_sentiment_polarity
count	35103.000000	35103.000000	35103.000000	35103.000000	35103.000000	35103.000000	35103.000000	35103.000000	35103.000000
mean	0.118899	0.352883	0.095213	0.755122	-0.257774	-0.519271	-0.106909	0.068607	0.151987
std	0.095983	0.102511	0.070615	0.245383	0.126191	0.289376	0.094034	0.260124	0.221970
min	-0.393750	0.000000	0.000000	0.000000	-1.000000	-1.000000	-1.000000	-1.000000	0.000000
25%	0.057623	0.305553	0.050000	0.600000	-0.325469	-0.700000	-0.125000	0.000000	0.000000
50%	0.118554	0.357435	0.100000	0.800000	-0.251157	-0.500000	-0.100000	0.000000	0.000000
75%	0.177140	0.409334	0.100000	1.000000	-0.185542	-0.300000	-0.050000	0.136364	0.250000
max	0.727841	1.000000	1.000000	1.000000	0.000000	0.000000	0.000000	1.000000	1.000000

	Model	Accuracy	F1_Score(training)	F1_Score(test)	Cross-Validation
0	Decision Tree Classification	0.605426	1.000000	0.605426	0.606602
1	K Nearest Neighbors Classification	0.677768	0.696511	0.677768	0.672026
2	Naive Bayes	0.679974	0.660301	0.679974	0.654819
3	Random Forest Classification	0.705117	1.000000	0.705117	0.693392

	url	timedelta	n_tokens_title	n_tokens_content	n_unique_tokens	n_non_stop_words	n_non_stop_unique_tokens	num_hrefs	num_self_hrefs	num_imgs	...	min_positive_polarity	max_positive_polarity	avg_negative_polarity	min_negative_polarity	max_negative_polarity	title_subjectivity	title_sentiment_polarity	abs_title_subjectivity	abs_title_sentiment_polarity	shares
0	http://mashable.com/2013/01/07/amazon-instant-...	731.0	12.0	219.0	0.663594	1.0	0.815385	4.0	2.0	1.0	...	0.100000	0.7	-0.350000	-0.600	-0.200000	0.500000	-0.187500	0.000000	0.187500	593
1	http://mashable.com/2013/01/07/ap-samsung-spon...	731.0	9.0	255.0	0.604743	1.0	0.791946	3.0	1.0	1.0	...	0.033333	0.7	-0.118750	-0.125	-0.100000	0.000000	0.000000	0.500000	0.000000	711
2	http://mashable.com/2013/01/07/apple-40-billio...	731.0	9.0	211.0	0.575130	1.0	0.663866	3.0	1.0	1.0	...	0.100000	1.0	-0.466667	-0.800	-0.133333	0.000000	0.000000	0.500000	0.000000	1500
3	http://mashable.com/2013/01/07/astronaut-notre...	731.0	9.0	531.0	0.503788	1.0	0.665635	9.0	0.0	1.0	...	0.136364	0.8	-0.369697	-0.600	-0.166667	0.000000	0.000000	0.500000	0.000000	1200
4	http://mashable.com/2013/01/07/att-u-verse-apps/	731.0	13.0	1072.0	0.415646	1.0	0.540890	19.0	19.0	20.0	...	0.033333	1.0	-0.220192	-0.500	-0.050000	0.454545	0.136364	0.045455	0.136364	505

	data_channel_is_lifestyle	data_channel_is_entertainment	data_channel_is_bus	data_channel_is_socmed	data_channel_is_tech	data_channel_is_world	weekday_is_monday	weekday_is_tuesday	weekday_is_wednesday	weekday_is_thursday	weekday_is_friday	weekday_is_saturday	weekday_is_sunday	is_weekend
count	39644.000000	39644.000000	39644.000000	39644.000000	39644.000000	39644.000000	39644.000000	39644.000000	39644.000000	39644.000000	39644.000000	39644.000000	39644.000000	39644.000000
mean	0.052946	0.178009	0.157855	0.058597	0.185299	0.212567	0.168020	0.186409	0.187544	0.183306	0.143805	0.061876	0.069039	0.130915
std	0.223929	0.382525	0.364610	0.234871	0.388545	0.409129	0.373889	0.389441	0.390353	0.386922	0.350896	0.240933	0.253524	0.337312
min	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000
25%	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000
50%	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000
75%	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000
max	1.000000	1.000000	1.000000	1.000000	1.000000	1.000000	1.000000	1.000000	1.000000	1.000000	1.000000	1.000000	1.000000	1.000000

	data_channel_is_entertainment	data_channel_is_bus	data_channel_is_tech	LDA_00	LDA_01	LDA_02	LDA_03	LDA_04
0	1.0	0.0	0.0	0.500331	0.378279	0.040005	0.041263	0.040123
1	0.0	1.0	0.0	0.799756	0.050047	0.050096	0.050101	0.050001
2	0.0	1.0	0.0	0.217792	0.033334	0.033351	0.033334	0.682188
3	1.0	0.0	0.0	0.028573	0.419300	0.494651	0.028905	0.028572
4	0.0	0.0	1.0	0.028633	0.028794	0.028575	0.028572	0.885427

	data_channel_is_lifestyle	data_channel_is_entertainment	data_channel_is_bus	data_channel_is_tech	data_channel_is_world	theme
0	0.0	1.0	0.0	0.0	0.0	Entertainment
1	0.0	0.0	1.0	0.0	0.0	Bus
2	0.0	0.0	1.0	0.0	0.0	Bus
3	0.0	1.0	0.0	0.0	0.0	Entertainment
4	0.0	0.0	0.0	1.0	0.0	Tech
5	0.0	0.0	0.0	1.0	0.0	Tech
6	1.0	0.0	0.0	0.0	0.0	Lifestyle
7	0.0	0.0	0.0	1.0	0.0	Tech
8	0.0	0.0	0.0	1.0	0.0	Tech
9	0.0	0.0	0.0	0.0	1.0	World

	Model	RMSE	R2_Score(training)	R2_Score(test)	Cross-Validation
0	Linear Regression	967.485513	0.071631	0.079926	0.065166
1	Lasso Regression	967.534244	0.071587	0.079833	0.065326
2	Decision Tree Regression	1355.454052	1.000000	-0.805941	-0.810282
3	Random Forest Regression	956.789514	0.875817	0.100157	0.095825
4	Support Vector Regression	1048.275079	-0.092854	-0.080151	-0.097776

	Model	Accuracy	F1_Score(training)	F1_Score(test)	Cross-Validation
0	Decision Tree Classification	0.251654	1.000000	0.251654	0.256122
1	K Nearest Neighbors Classification	0.275915	0.366550	0.275915	0.278813
2	Naive Bayes	0.301500	0.298572	0.301500	0.293088
3	Random Forest Classification	0.326643	1.000000	0.326643	0.329206

	Actual shares	Predicted shares
0	1000	1636.53
1	723	1772.85
2	684	1007.76
3	1300	1240.48
4	1100	2044.48

Introduction¶

Description and goal of the project¶

Description of the dataset¶

Importations¶

Dataset Downloading & Characteristics¶

Data Cleaning¶

Shares column (our target)¶

Using Quartiles¶

Using IQR Rule¶

Using Z-Score¶

IQR Rule Method¶

Drop column method¶

Token Columns¶

Focus on 'n_non_stop_words'¶

Return to token columns¶

Videos & Images¶

Webscraping¶

LDAs and Data channels¶

Keywords Columns¶

References Columns¶

Remarging the "weekday" columns¶

Positive/Negative Words Rate Columns¶

Polarity Columns¶

Subjectivity Columns¶

Creating two DataFrame : capped one & cleaned one¶

Correlations of each features with 'shares' (target) for both DataFrames¶

Correlations Matrixes of both DataFrames¶

Boxplots of each features for both DataFrames¶

Predictive Models with 'data_quality'¶

Regression¶

LinearRegression¶

LassoRegressor¶

DecisionTreeRegressor¶

RandomForestRegressor¶

SupportVectorRegressor¶

Evaluation of models¶

Analysis of the models¶

Features Importances & Performance Visualisation of Random Forest Regressor¶

Classification¶

Binary Classification (Not popular/Popular)¶

DecisionTreeClassifier¶

K Nearest Neighbors¶

Naive Bayes Classifier¶

RandomForestClassifier¶

Evaluation of models¶

Analysis of the models¶

GridSearch¶

Testing with Top 15 features¶

Multi-Labels¶

Deep Learning using TensorFlow¶

Binary classification¶

Multi-labels classification¶

Total results¶

Data analysis¶

I) Title and content words¶

II) Sentiment and subjectivity¶

III) Day of publication¶

IV) Data channel¶

V) Images & Videos¶

VI) References¶

Conclusion¶